{ "pages": [ { "page_number": 1, "text": "" }, { "page_number": 2, "text": "" }, { "page_number": 3, "text": "" }, { "page_number": 4, "text": "A N A U E R B A C H B O O K\nCRC Press is an imprint of the\nTaylor & Francis Group, an informa business\nBoca Raton London New York\n" }, { "page_number": 5, "text": "Auerbach Publications\nTaylor & Francis Group\n6000 Broken Sound Parkway NW, Suite 300\nBoca Raton, FL 33487-2742\n© 2009 by Taylor & Francis Group, LLC \nAuerbach is an imprint of Taylor & Francis Group, an Informa business\nNo claim to original U.S. Government works\nPrinted in the United States of America on acid-free paper\n10 9 8 7 6 5 4 3 2 1\nInternational Standard Book Number-13: 978-0-8493-8250-5 (Hardcover)\nThis book contains information obtained from authentic and highly regarded sources. Reasonable \nefforts have been made to publish reliable data and information, but the author and publisher can-\nnot assume responsibility for the validity of all materials or the consequences of their use. The \nauthors and publishers have attempted to trace the copyright holders of all material reproduced \nin this publication and apologize to copyright holders if permission to publish in this form has not \nbeen obtained. If any copyright material has not been acknowledged please write and let us know so \nwe may rectify in any future reprint.\nExcept as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, \ntransmitted, or utilized in any form by any electronic, mechanical, or other means, now known or \nhereafter invented, including photocopying, microfilming, and recording, or in any information \nstorage or retrieval system, without written permission from the publishers.\nFor permission to photocopy or use material electronically from this work, please access www.copy-\nright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 \nRosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-\nvides licenses and registration for a variety of users. For organizations that have been granted a \nphotocopy license by the CCC, a separate system of payment has been arranged.\nTrademark Notice: Product or corporate names may be trademarks or registered trademarks, and \nare used only for identification and explanation without intent to infringe.\nLibrary of Congress Cataloging-in-Publication Data\nZhang, Yan, 1977-\nSecurity in wireless mesh networks / Yan Zhang, Jun Zheng, and Honglin Hu.\np. cm.\nIncludes bibliographical references and index.\nISBN 978-0-8493-8250-5 (alk. paper)\n1. Wireless communication systems--Security measures. 2. Computer \nnetworks--Security measures. 3. Routers (Computer networks) I. Zheng, Jun, \nPh.D. II. Hu, Honglin, 1975- III. Title. \nTK5103.2.Z53 2007\n005.8--dc22\n2007011243\nVisit the Taylor & Francis Web site at\nhttp://www.taylorandfrancis.com\nand the Auerbach Web site at\nhttp://www.auerbach-publications.com\n" }, { "page_number": 6, "text": "Contents\nContributors\n. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii\nPART I: INTRODUCTION\n1\nAn Introduction\nto Wireless Mesh Networks . . . . . . . . . . . . . . . . . . .3\nA. Antony Franklin and C. Siva Ram Murthy\n2\nMesh Networking in Wireless PANs, LANs,MANs,\nand WANs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45\nNeila Krichene and Noureddine Boudriga\nPART II: SECURITY PROTOCOLS AND TECHNIQUES\n3\nAttacks and Security Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 111\nAnjum Naveed, Salil S. Kanhere, and Sanjay K. Jha\n4\nIntrusion\nDetection in Wireless Mesh Networks . . . . . . . . . . . . 145\nThomas M. Chen, Geng-Sheng Kuo, Zheng-Ping Li,\nand Guo-Mei Zhu\n5\nSecure Routing in Wireless Mesh Networks . . . . . . . . . . . . . . . . . 171\nManel Guerrero Zapata\n6\nHop Integrity in Wireless Mesh Networks . . . . . . . . . . . . . . . . . . . 197\nChin-Tser Huang\n7\nPrivacy Preservation\nin Wireless Mesh Networks . . . . . . . . . . . 227\nTaojun Wu, Yuan Xue, and Yi Cui\n8\nProviding Authentication,\nTrust, and Privacy in\nWireless Mesh Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261\nHassnaa Moustafa\nv\n" }, { "page_number": 7, "text": "vi\n■\nContents\n9\nNon-Interactive\nKey Establishment\nin Wireless\nMesh Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .297\nZhenjiang Li and J.J. Garcia-Luna-Aceves\n10\nKey Management in Wireless Mesh Networks . . . . . . . . . . . . . . .323\nManel Guerrero Zapata\nPART III: SECURITY STANDARDS, APPLICATIONS,\nAND ENABLING TECHNOLOGIES\n11\nSecurity in Wireless PANMesh Networks . . . . . . . . . . . . . . . . . . . .349\nStefaan Seys, Dave Singel´ee, and Bart Preneel\n12\nSecurity in Wireless LANMesh Networks . . . . . . . . . . . . . . . . . . . .381\nNancy-Cam Winget and Shah Rahman\n13\nSecurity in IEEE802.15.4\nCluster-Based Networks . . . . . . . . . . 409\nMoazzam Khan and Jelena Misic\n14\nSecurity in Wireless Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . 433\nYong Wang, Garhan Attebury, and Byrav Ramamurthy\n15\nKey Management in Wireless Sensor Networks . . . . . . . . . . . . . 491\nFalko Dressler\nIndex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517\n" }, { "page_number": 8, "text": "List of Contributors\nGarhan Attebury\nUniversity of Nebraska-Lincoln\nLincoln, Nebraska\nNoureddine Boudriga\nCNAS Research Lab\nUniversity of Carthage\nCarthage, Tunisia\nThomas M. Chen\nSouthern Methodist University\nDallas, Texas\nYi Cui\nDepartment of Electrical Engineering\nand Computer Science\nVanderbilt University\nNashville, Tennessee\nFalko Dressler\nAutonomic Networking Group\nDepartment of Computer Sciences\nUniversity of Erlangen\nNuremberg, Germany\nA. Antony Franklin\nIndian Institute of\nTechnology Madras\nChennai, Tamilnadu, India\nJ.J. Garcia-Luna-Aceves\nComputer Engineering\nUniversity of California\nSanta Cruz, California\nChin-Tser Huang\nUniversity of South Carolina\nColumbia, South Carolina\nSanjay K. Jha\nSchool of Computer Science\nand Engineering\nUniversity of New South Wales\nSydney, Australia\nSalil S. Kanhere\nSchool of Computer Science\nand Engineering\nUniversity of New South Wales\nSydney, Australia\nvii\n" }, { "page_number": 9, "text": "viii\n■\nContributors\nMoazzam Khan\nManitoba University\nManitoba, Winnipeg, Canada\nNeila Krichene\nCNAS Research Lab\nUniversity of Carthage\nCarthage, Tunisia\nGeng-Sheng Kuo\nBeijing University of Posts\nand Telecommunications\nBeijing, China\nZhenjiang Li\nComputer Engineering,\nUniversity of California, Santa Cruz\nSanta Cruz, California\nZheng-Ping Li\nBeijing University of Posts\nand Telecommunications\nBeijing, China\nJelena Misic\nManitoba University\nManitoba, Winnipeg, Canada\nHassnaa Moustafa\nFrance Telecom R&D\nParis, France\nC. Siva Ram Murthy\nIndian Institute of\nTechnology Madras\nChennai, Tamilnadu, India\nAnjum Naveed\nSchool of Computer Science\nand Engineering\nUniversity of New South Wales\nSydney, Australia\nBart Preneel\nDepartment of Electrical\nEngineering\nKatholieke Universiteit\nLeuven, Belgium\nShah Rahman\nCisco Systems\nSan Jose, California\nByrav Ramamurthy\nUniversity of Nebraska-Lincoln\nLincoln, Nebraska\nStefaan Seys\nDepartment of Electrical Engineering\nKatholieke Universiteit\nLeuven, Belgium\nDave Singel ´ee\nDepartment of Electrical\nEngineering\nKatholieke Universiteit\nLeuven, Belgium\nYong Wang\nUniversity of Nebraska-Lincoln\nLincoln, Nebraska\nNancy-Cam Winget\nCisco Systems\nSan Jose, California\n" }, { "page_number": 10, "text": "Contributors\n■\nix\nTaojun Wu\nDepartment of Electrical Engineering\nand Computer Science\nVanderbilt University\nNashville, Tennessee\nYuan Xue\nDepartment of Electrical Engineering\nand Computer Science\nVanderbilt University\nNashville, Tennessee\nManel Guerrero\nZapata\nTechnical University\nof Catalonia\nBarcelona, Spain\nGuo-Mei Zhu\nBeijing University of Posts\nand Telecommunications\nBeijing, China\n" }, { "page_number": 11, "text": "" }, { "page_number": 12, "text": "INTRODUCTION\nI\n" }, { "page_number": 13, "text": "" }, { "page_number": 14, "text": "Chapter 1\nAn Introduction to\nWireless Mesh Networks\nA. Antony Franklin and C. Siva Ram Murthy\nContents\n1.1\nIntroduction ............................................................5\n1.1.1\nSingle-Hop and Multi-Hop Wireless Networks ...............6\n1.1.2\nAd hoc Networks and WMNs .................................7\n1.2\nArchitecture of WMNs..................................................8\n1.3\nApplications of WMNs .................................................9\n1.4\nIssues in WMNs .......................................................13\n1.4.1\nCapacity .......................................................14\n1.4.2\nPhysical Layer.................................................15\n1.4.3\nMedium Access Scheme ......................................17\n1.4.4\nRouting........................................................20\n1.4.4.1\nRouting Metrics for WMNs ..........................20\n1.4.4.2\nRouting Protocols for WMNs........................22\n1.4.5\nTransport Layer ...............................................23\n1.4.6\nGateway Load Balancing .....................................24\n1.4.7\nSecurity .......................................................26\n1.4.8\nPower Management ..........................................27\n1.4.9\nMobility Management ........................................28\n1.4.10\nAdaptive Support for Mesh Routers and Mesh Clients ......29\n3\n" }, { "page_number": 15, "text": "4\n■\nSecurity in Wireless Mesh Networks\n1.4.11\nIntegration with Other Network Technologies ..............30\n1.4.12\nDeployment Considerations..................................31\n1.5\nWMN Deployments/Testbeds .........................................34\n1.5.1\nIEEE 802.11 WMNs ...........................................34\n1.5.2\nIEEE 802.15 WMNs ...........................................35\n1.5.3\nIEEE 802.16 WMNs ...........................................36\n1.5.4\nAcademic Research Testbeds.................................37\n1.5.5\nIndustrial Research in WMNs ................................38\n1.5.6\nMesh Networking Products ..................................39\n1.6\nSummary...............................................................40\nReferences...................................................................41\nWireless mesh networking has emerged as a promising concept to meet\nthe challenges in next-generation wireless networks such as providing flex-\nible, adaptive, and reconfigurable architecture while offering cost-effective\nsolutions to service providers. Several architectures for wireless mesh net-\nworks (WMNs) have been proposed based on their applications [1]. One of\nthe most general forms of WMNs interconnects the stationary and mobile\nclients to the Internet efficiently by the core nodes in multi-hop fashion.\nThe core nodes are the mesh routers which form a wireless mesh back-\nbone among them. The mesh routers provide a rich radio mesh connectivity\nwhich significantly reduces the up-front deployment cost and subsequent\nmaintenance cost. They have limited mobility and forward the packets re-\nceived from the clients to the gateway router which is connected to the\nbackhaul network/Internet. The mesh backbone formed by mesh routers\nprovides a high level of reliability. WMNs are being considered for a wide\nvariety of applications such as backhaul connectivity for cellular radio ac-\ncess networks, high-speed metropolitan area mobile networks, community\nnetworking, building automation, intelligent transport system networks, de-\nfense systems, and citywide surveillance systems. Prior efforts on wireless\nnetworks, especially multi-hop ad hoc networks, have led to significant\nresearch contributions that range from fundamental results on theoretical\ncapacity bounds to development of efficient routing and transport layer\nprotocols. However, the recent work is on deploying sizable WMNs and\nother important aspects such as network radio range, network capacity,\nscalability, manageability, and security. There are a number of research is-\nsues in different layers of the protocol stack and a number of standards\nare coming up for the implementation of WMNs for WANs, MANs, LANs,\nand PANs. The mesh networking testbeds by industries and academia fur-\nther enhanced the research in WMNs. The mesh networking products by\ndifferent vendors are making WMNs a reality.\n" }, { "page_number": 16, "text": "An Introduction to Wireless Mesh Networks\n■\n5\nInternet\nInternet\nMesh router\nClient node\nWireless link\nGateway\nCellular network\nSensor network\nWLAN\nPDA\nEdge router\nEdge router\nEdge router\nEdge router\nWired backbone link\nEdge\nrouter\nGateway\nFigure 1.1\nArchitecture of a wireless mesh network.\n1.1\nIntroduction\nWMNs are multi-hop wireless networks formed by mesh routers and mesh\nclients. These networks typically have a high data rate and low deployment\nand maintenance overhead. Mesh routers are typically stationary and do not\nhave energy constraints, but the clients are mobile and energy constrained.\nSome mesh routers are designated as gateway routers which are connected\nto the Internet through a wired backbone. A gateway router provides ac-\ncess to conventional clients and interconnects ad hoc, sensor, cellular, and\nother networks to the Internet, as shown in Figure 1.1. A mesh network can\nprovide multi-hop communication paths between wireless clients, thereby\nserving as a community network, or can provide multi-hop paths between\nthe client and the gateway router, thereby providing broadband Internet\naccess to clients. As there is no wired infrastructure to deploy in the case\nof WMNs, they are considered cost-effective alternatives to WLANs (wire-\nless local area networks) and backbone networks to mobile clients. The\n" }, { "page_number": 17, "text": "6\n■\nSecurity in Wireless Mesh Networks\nexisting wireless networking technologies such as IEEE 802.11, IEEE 802.15,\nIEEE 802.16, and IEEE 802.20 are used in the implementation of WMNs. The\nIEEE 802.11 is a set of WLAN standards that define many aspects of wireless\nnetworking. One such aspect is mesh networking, which is currently un-\nder development by the IEEE 802.11 Task Group. Recently, there has been\ngrowing research and practical interest in WMNs. There are numerous on-\ngoing projects on wireless mesh networks in academia, research labs, and\ncompanies. Many academic institutions developed their own testbed for\nresearch purposes. These efforts are toward developing various applica-\ntions of WMNs such as home, enterprise, and community networking. As\nthe WMNs use multi-hop paths between client nodes or between a client\nand a gateway router, the existing protocols for multi-hop ad hoc wireless\nnetworks are well suited for WMNs. The ongoing work in WMNs is on\nincreasing the throughput and developing efficient protocols by utilizing\nthe static nature of the mesh routers and topology.\n1.1.1\nSingle-Hop and Multi-Hop Wireless Networks\nGenerally, wireless networks are classified as single-hop and multi-hop\nnetworks. In a single-hop network, the client connects to the fixed base\nstation or access point directly in one hop. The well-known examples of\nsingle-hop wireless networks are WLANs and cellular networks. WLANs\ncontain special nodes called access points (APs), which are connected to\nexisting wired networks such as Ethernet LANs. The mobile devices are\nconnected to the AP through a one-hop wireless link. Any communication\nbetween mobile devices happens via AP. In the case of cellular networks,\nthe geographical area to be covered is divided into cells which are usually\nconsidered to be hexagonal. A base station (BS) is located in the center of\nthe cell and the mobile terminals in that cell communicate with it in a single-\nhop fashion. Communication between any two mobile terminals happens\nthrough one or more BSs. These networks are called infrastructure wireless\nnetworks because they are infrastructure (BS) dependent. The path setup\nbetween two clients (mobile nodes), say node A and node B, is completed\nthrough the BS, as shown in Figure 1.2.\nIn a multi-hop wireless network, the source and destination nodes com-\nmunicate in a multi-hop fashion. The packets from the source node traverse\nthrough one or more intermediate/relaying nodes to reach the destination.\nBecause all nodes in the network also act as routers, there is no need\nfor a BS or any other dedicated infrastructure. Hence, such networks are\nalso called infrastructure-less networks. The well-known forms of multi-hop\nnetworks are ad hoc networks, sensor networks, and WMNs. Communica-\ntion between two nodes, say node C and node F, takes place through the\nrelaying nodes D and E, as shown in Figure 1.3.\n" }, { "page_number": 18, "text": "An Introduction to Wireless Mesh Networks\n■\n7\nE\nD\nSwitching center\n+\nGateway\nC\nA\nB\nCommunication path\nBase station\nMobile node\nFigure 1.2\nSingle-hop network scenario (cellular network).\nIn the case of single-hop networks, complete information about the\nclients is available at the BS and the routing decisions are made in a cen-\ntralized fashion, thus making routing and resource management simple.\nBut it is not the case in multi-hop networks. All the mobile nodes have to\ncoordinate among themselves for communication between any two nodes.\nHence, routing and resource management are done in a distributed way.\n1.1.2\nAd hoc Networks and WMNs\nIn ad hoc networks, all the nodes are assumed to be mobile and there is\nno fixed infrastructure for the network. These networks find applications\nwhere fixed infrastructure is not possible, such as military operations in\nthe battlefield, emergency operations, and networks of handheld devices.\nBecause of lack of infrastructure the nodes have to cooperate among them-\nselves to form a network. Due to mobility of the nodes in the network, the\nnetwork topology changes frequently. So the protocols for ad hoc networks\nhave to handle frequent changes in the topology. In most of the applica-\ntions of ad hoc networks, the mobile devices are energy constrained as\n" }, { "page_number": 19, "text": "8\n■\nSecurity in Wireless Mesh Networks\nB\nA\nE\nF\nC\nD\nMobile node\nWireless link\nCommunication path\nFigure 1.3\nMulti-hop network scenario (ad hoc network).\nthey are operating on battery. This requires energy-efficient networking\nsolutions for ad hoc networks. But in the case of WMNs, mesh routers are\nassumed to be fixed (or have limited mobility) and form a fixed mesh infra-\nstructure. The clients are mobile or fixed and utilize the mesh routers to\ncommunicate to the backhaul network through the gateway routers and to\nother clients by using mesh routers as relaying nodes. These networks find\napplications where networks of fixed wireless nodes are necessary. There\nare several architectures for mesh networks, depending on their applica-\ntions. In the case of infrastructure backbone networking, the edge routers\nare used to connect different networks to the mesh backbone and the inter-\nmediate routers are used as multi-hop relaying nodes to the gateway router,\nas shown in Figure 1.1. But in the case of community networking, every\nrouter provides access to clients and also acts as a relaying node between\nmesh routers.\n1.2\nArchitecture of WMNs\nThere are two types of nodes in a WMN called mesh routers and mesh\nclients. Compared to conventional wireless routers that perform only\nrouting, mesh routers have additional functionalities to enable mesh\n" }, { "page_number": 20, "text": "An Introduction to Wireless Mesh Networks\n■\n9\nnetworking. The mesh routers have multiple interfaces of the same or\ndifferent communications technologies based on the requirement. They\nachieve more coverage with the same transmission power by using multi-\nhop communication through other mesh routers. They can be built on\ngeneral-purpose computer systems such as PCs and laptops, or can be built\non dedicated hardware platforms (embedded systems). There are a vari-\nety of mesh clients such as laptop, desktop, pocket PCs, IP phones, RFID\nreaders, and PDAs. The mesh clients have mesh networking capabilities to\ninteract with mesh routers, but they are simpler in hardware and software\ncompared to mesh routers. Normally they have a single communication\ninterface built on them. The architecture of WMNs (shown in Figure 1.1)\nis the most common architecture used in many mesh networking appli-\ncations such as community networking and home networking. The mesh\nrouters shown have multiple interfaces with different networking technolo-\ngies which provide connectivity to mesh clients and other networks such as\ncellular and sensor networks. Normally, long-range communication tech-\nniques such as directional antennas are provided for communication be-\ntween mesh routers. Mesh routers form a wireless mesh topology that has\nself-configuration and self-healing functions built into them. Some mesh\nrouters are designated as gateways which have wired connectivity to the\nInternet. The integration of other networking technologies is provided by\nconnecting the BS of the network that connects to WMNs to the mesh\nrouters. Here, the clients communicate to the BS of its own network and\nthe BS in turn communicates to the mesh router to access the WMN.\n1.3\nApplications of WMNs\nWMNs introduce the concept of a peer-to-peer mesh topology with wire-\nless communication between mesh routers. This concept helps to overcome\nmany of today’s deployment challenges, such as the installation of exten-\nsive Ethernet cabling, and enables new deployment models. Deployment\nscenarios that are particularly well suited for WMNs include the following:\n■\nCampus environments (enterprises and universities), manufacturing,\nshopping centers, airports, sporting venues, and special events\n■\nMilitary operations, disaster recovery, temporary installations, and\npublic safety\n■\nMunicipalities, including downtown cores, residential areas, and\nparks\n■\nCarrier-managed service in public areas or residential communities\nDue to the recent research advances in WMNs, they have been used in\nnumerous applications. The mesh topology of the WMNs provides many\n" }, { "page_number": 21, "text": "10\n■\nSecurity in Wireless Mesh Networks\nalternative paths for any pair of source and destination nodes, resulting in\nquick reconfiguration of the path when there is a path failure. WMNs pro-\nvide the most economical data transfer coupled with freedom of mobility.\nMesh routers can be placed anywhere such as on the rooftop of a home\nor on a lamppost to provide connectivity to mobile/static clients. Mesh\nrouters can be added incrementally to improve the coverage area. These\nfeatures of WMNs attract the research community to use WMNs in different\napplications:\n■\nHome Networking: Broadband home networking is a network of\nhome appliances (personal computer, television, video recorder,\nvideo camera, washing machine, refrigerator) realized by WLAN\ntechnology. The obvious problem here is the location of the access\npoint in the home, which may lead to dead zones without service\ncoverage. More coverage can be achieved by multiple access points\nconnected using Ethernet cabling, which leads to an increase in\ndeployment cost and overhead. These problems can be solved by\nreplacing all the access points by the mesh routers and establishing\nmesh connectivity between them. This provides broadband con-\nnectivity between the home networking devices and only a single\nconnection to the Internet is needed through the gateway router. By\nchanging the location and number of mesh routers, the dead zones\ncan be eliminated. Figure 1.4 shows a typical home network using\nmesh routers.\n■\nCommunity and Neighborhood Networking: The usual way of estab-\nlishing community networking is connecting the home network/PC\nto the Internet with a cable or DSL modem. All the traffic in commu-\nnity networking goes through the Internet, which leads to inefficient\nutilization of the network resources. The last mile of wireless con-\nnectivity might not provide coverage outside the home. Community\nnetworking by WMNs solves all these problems and provides a cost-\neffective way to share Internet access and other network resources\namong different homes. Figure 1.5 shows wireless mesh network-\ning by placing the mesh routers on the rooftop of houses. There are\nmany advantages to enabling such mesh connectivity to form a com-\nmunity mesh network. For example, when enough neighbors coop-\nerate and forward each others’ packets, they do not need individual\nInternet connectivity; instead they can get faster, cost-effective Inter-\nnet access via gateways distributed in their neighborhood. Packets\ndynamically find a route, hopping from one neighbor’s node to an-\nother to reach the Internet through one of these gateways. Another\nadvantage is that neighbors can cooperatively deploy backup tech-\nnology and never have to worry about losing information due to a\n" }, { "page_number": 22, "text": "An Introduction to Wireless Mesh Networks\n■\n11\nLaptop\nTelephone\nPrinter\nPDA\nMesh router\nTV\nMesh router\nMesh router\nMesh router\nMesh router\nPDA\nMesh router\nWireless link between client and mesh router\nDesktop\nMesh router\nWireless link between mesh routers\nCamcorder\nFigure 1.4\nWireless mesh network-based home networking.\ncatastrophic disk failure. Another advantage is that this technology\nalleviates the need for routing traffic belonging to community net-\nworking through the Internet. For example, distributed file storage,\ndistributed file access, and video streaming applications in the com-\nmunity share network resources in the WMNs without using the\nInternet, which improves the performance of these applications.\nNeighborhood community networks allow faster and easier dissemi-\nnation of cached information that is relevant to the local community.\nMesh routers can be easily mounted on rooftops or windows and\nthe client devices get connected to them in a single hop.\n■\nSecurity Surveillance System: As security is turning out to be of very\nhigh concern, security surveillance systems are becoming a necessity\nfor enterprise buildings and shopping malls. The security surveill-\nance system needs high bandwidth and a reliable backbone network\nto communicate surveillance information, such as images, audio, and\n" }, { "page_number": 23, "text": "12\n■\nSecurity in Wireless Mesh Networks\nInternet\nHome with rooftop mesh router\nWireless link between mesh routers\nWired backbone connectivity\nGateway\nFigure 1.5\nWireless mesh network-based community networking.\nvideo, and low-cost connectivity between the surveillance devices.\nThe recent advances of WMNs provide high bandwidth and reliable\nbackbone connectivity and an easy way of connecting surveillance\ndevices located in different places with low cost.\n■\nDisaster Management and Rescue Operations: WMNs can be used\nin places where spontaneous network connectivity is required, such\nas disaster management and emergency operations. During disasters\nlike fire, flood, and earthquake, all the existing communication in-\nfrastructures might be collapsed. So during the rescue operation,\nmesh routers can be placed at the rescue team vehicle and different\nlocations which form the high-bandwidth mesh backbone network,\nas shown in Figure 1.6. This helps rescue team members to com-\nmunicate with each other. By providing different communication\n" }, { "page_number": 24, "text": "An Introduction to Wireless Mesh Networks\n■\n13\nWireless link between mesh routers\nWireless link between mobile terminal and mesh router\nMobile terminal with rescue team\nRescue vehicle\nFigure 1.6\nWireless mesh network-based rescue operation.\ninterfaces at the mesh routers, different mobile devices get access to\nthe network. This helps people to communicate with others when\nthey are in critical situations. These networks can be established in\nless time, which makes the rescue operation more effective.\n1.4\nIssues in WMNs\nVarious research issues in WMNs are described in this section. As WMNs\nare also multi-hop wireless networks like ad hoc networks, the protocols\ndeveloped for ad hoc networks work well for WMNs. Many challenging\nissues in ad hoc networks have been addressed in recent years. WMNs\nhave inherent characteristics such as a fixed mesh backbone formed by\nmesh routers, resource-rich mesh routers, and resource-constrained clients\n" }, { "page_number": 25, "text": "14\n■\nSecurity in Wireless Mesh Networks\ncompared to ad hoc networks. Due to this, WMNs require considerable\nwork to address the problems that arise in each layer of the protocol stack\nand system implementation.\n1.4.1\nCapacity\nThe primary concern of WMNs is to provide high-bandwidth connectivity to\ncommunity and enterprise users. In a single-channel wireless network, the\ncapacity of the network degrades as the number of hops or the diameter\nof the network increases due to interference. The capacity of the WMN\nis affected by many factors such as network architecture, node density,\nnumber of channels used, node mobility, traffic pattern, and transmission\nrange. A clear understanding of the effect of these factors on capacity of\nthe WMNs provides insight to protocol design, architecture design, and\ndeployment of WMNs.\nIn [2] Gupta and Kumar analytically studied the upper and lower bounds\nof the capacity of wireless ad hoc networks. They showed that the through-\nput capacity of the nodes reduces significantly when node density increa-\nses. The maximum achievable throughput of randomly placed n identical\nnodes each with a capacity of W bits/second is \u0002(\nW\n√\nn∗log(n)) bits/second\nunder a non-interference protocol. Even under optimal circumstances the\nmaximum achievable throughput is only \u0002( W\n√n) bits/second. The capacity\nof the network can be increased by deploying relaying nodes and using a\nmulti-hop path for transmission.\nThe IEEE 802.11 standard [4] provides a number of channels in the\navailable radio spectrum, but some of them may be interfering with each\nother. If the interfering channels are used simultaneously, then the data\ngets corrupted at the receiving end. But the non-overlapping channels can\nbe used simultaneously by different nodes in the same transmission range\nwithout any collision of the data. IEEE 802.11b [6] provides 3 such non-\noverlapping channels at 2.4 GHz band and IEEE 802.11a [5] provides 13\nnon-overlapping channels at 5 GHz band. These orthogonal channels can\nbe used simultaneously at different nodes in the network to improve the\ncapacity of the network. In multi-channel multi-radio communication each\nnode is provided with more than one radio interface (say m) and each\ninterface is assigned one of the orthogonal channels available (say n). If\neach node has n number of radio interfaces (m = n) and each orthogonal\nchannel is assigned to one interface, then the network can achieve n-fold\nincrease in capacity because the n interfaces can transmit simultaneously\nwithout any interference with each other. But normally the number of in-\nterfaces is less than the number of available channels (m < n) due to the\ncost of the interfaces and the complexity of the nodes. In this case an m-\nfold increase in capacity can be achieved by assigning m interfaces with m\n" }, { "page_number": 26, "text": "An Introduction to Wireless Mesh Networks\n■\n15\ndifferent orthogonal channels. Moreover, when m < n the capacity bound\nof a multi-channel multi-radio wireless mesh network depends on the ratio\nof n and m [7].\n1.4.2\nPhysical Layer\nThe network capacity mainly depends on the physical layer technique used.\nThere are a number of physical layer techniques available with different\noperating frequencies and they provide different transport capacity in wire-\nless communications. Some existing wireless radios even provide multi-\nple transmission rates by different combinations of modulation and coding\ntechniques [6]. In such networks, the transmission rate is chosen by link\nadaptation techniques. Normally, link signal-to-noise ratio (SNR) or carrier-\nto-noise ratio (CNR) from the physical layer is considered for link adapta-\ntion, but this alone does not describe the signal quality in the environment\nlike frequency-selective fading channel. To overcome the problems with RF\ntransmission, other physical layer techniques have been used for wireless\ncommunications. Some high-speed physical layer techniques are available\nwhich improve the capacity of the wireless networks significantly. Some of\nthe techniques for improving the capacity of WMNs are described in this\nsection.\n■\nOrthogonal Frequency Division Multiplexing (OFDM): The OFDM\ntechnique is based on the principle of Frequency Division Multi-\nplexing (FDM) with digital modulation schemes. The bit stream to\nbe transmitted is split into a number of parallel low bit rate streams.\nThe available frequency spectrum is divided into many sub-channels\nand each low bit rate stream is transmitted by modulating over a\nsub-channel using a standard modulation scheme such as Phase\nShift Keying (PSK) and Quadrature Amplitude Modulation (QAM).\nThe primary advantage of OFDM is its ability to work under severe\nchannel conditions, such as multi-path and narrow-band interfer-\nence, without complex equalization filters at the transmitter and re-\nceiver. The OFDM technique has increased the transmission rate of\nIEEE 802.11 networks from 11 to 54 Mbps.\n■\nUltra Wide Band (UWB): UWB technology provides much higher\ndata rate (ranges from 3 to 10 GHz) compared to other RF transmis-\nsion technologies. A significant difference between traditional radio\ntransmission and UWB radio transmission is that traditional radio\ntransmission transmits information by varying the power, frequency,\nor phase in distinct and controlled frequencies while UWB trans-\nmission transmits information by generating radio energy at specific\ntimes with a broad frequency range. Due to this, UWB transmission\n" }, { "page_number": 27, "text": "16\n■\nSecurity in Wireless Mesh Networks\nis immune to multi-path fading and interference,1 which are com-\nmon in any radio transmission technique. UWB wireless links have\nthe characteristic that the bandwidth decreases rapidly as the dis-\ntance increases. On the other hand UWB provides hundreds of non-\ninterfering channels within radio range of each other. Hence, UWB\nis applicable for only short-range communications such as WPAN.\nMesh architecture combined with UWB wireless technology allows\na very easy installation of communications infrastructure in offices\nor homes by deploying many repeater modules every 10 meters. As\nthese repeater modules require power to operate on, they have to\nbe placed with ceiling lights or floor power boxes. The IEEE 802.15\nTG4a standard for WPAN uses a UWB physical layer technique con-\nsisting of a UWB impulse radio (operating in unlicensed UWB spec-\ntrum) and a chirp spread spectrum (operating in unlicensed 2.4 GHz\nspectrum).\n■\nMultiple-Input Multiple-Output (MIMO): The use of multiple an-\ntennas at the transmitter and receiver, popularly known as MIMO\nwireless, is an emerging, cost-effective technology that makes high\nbandwidth wireless links a reality. MIMO significantly increases the\nthroughput and range with the same bandwidth and overall trans-\nmission power expenditure. This increase in throughput and range\nis by exploiting the multi-path propagation phenomena in wire-\nless communications. In general, the MIMO technique increases\nthe spectral efficiency of a wireless communications system. It has\nbeen shown by Telatar that the channel capacity (a theoretical up-\nper bound on system throughput) for a MIMO system increases as\nthe number of antennas increases, proportional to the minimum of\ntransmitter and receiver antennas [8]. MIMO can also be used in\nconjunction with OFDM and is part of the IEEE 802.16 standard.\n■\nSmart Antenna: The smart antenna technique improves the capacity\nof wireless networks by adding the directionality for transmission\nand reception of signals at the transmitter and receiver antenna.\nThis also helps in increasing energy efficiency. In cellular networks,\ndue to complexity and cost of smart antennas, it is implemented\nin BS alone. The directional antenna system is actively researched\nin ad hoc networks also. There are some directional antenna sys-\ntems available that can be tuned to certain directions by electronic\nbeam forming. This technique improves the performance of wireless\n1 In RF transmission, when the transmitted signal is reflected by mountains or buildings\nthe radio signal reaches the receiving antenna along two or more paths. The effect of\nthis multi-path reception includes constructive and destructive interference and phase\nshifting of the signal.\n" }, { "page_number": 28, "text": "An Introduction to Wireless Mesh Networks\n■\n17\nnetworks by reducing interference between the transmissions of dif-\nferent nodes in the network. But the use of a directional antenna\nnecessitates special MAC (Medium Access Control) protocols to sup-\nport directionality in transmission and reception.\n1.4.3\nMedium Access Scheme\nThe MAC (Medium Access Control) protocols for wireless networks are lim-\nited to single-hop communication while the routing protocols use multi-hop\ncommunication. The MAC protocols for WMNs are classified into single-\nchannel and multi-channel MAC. They are discussed in this section.\n■\nSingle-Channel MAC:\nThere are several MAC schemes which use\nsingle-channel for communication in the network. They are further\nclassified as (1) contention-based protocols, (2) contention-based\nprotocols with a reservation mechanism, and (3) contention-based\nprotocols with a scheduling mechanism.\n■\nContention-based protocols: These protocols have a contention-\nbased channel access policy among the nodes contending for\nthe channel. All the ready nodes in the network start contend-\ning for the channel simultaneously and the winning node gains\naccess to the channel. As the nodes cannot provide guaranteed\nbandwidth, these protocols cannot be used in carrying real-time\ntraffic, which requires QoS (quality of service) guarantees from\nthe system. Some of the contention-based protocols are MACAW\n(a media access protocol for Wireless LANs) [9], FAMA (Floor\nAcquisition Multiple Access protocol) [10], BTMA (Busy Tone\nMultiple Access protocol) [11], and MACA-BI (Multiple Access\nCollision Avoidance By Invitation) [12].\n■\nContention-based protocols with a reservation mechanism: Be-\ncause the contention-based protocols cannot provide guaran-\nteed access to the channel, they cannot be used in networks\nwhere real-time traffic has to be supported. To support real-\ntime traffic, some protocols reserve the bandwidth a priori. Such\nprotocols can provide QoS support for time-sensitive traffic.\nIn this type of protocol, the contention occurs during the re-\nsource (bandwidth) reservation phase. Once the bandwidth is\nreserved, the nodes get exclusive access to the reserved band-\nwidth. Hence, these protocols can provide QoS support for time-\nsensitive traffic. Some of the examples for these type of protocols\nare D-PRMA (Distributed Packet Reservation Multiple Access\nprotocol) [13], CATA (Collision Avoidance Time Allocation\n" }, { "page_number": 29, "text": "18\n■\nSecurity in Wireless Mesh Networks\nprotocol) [14], HRMA (Hop Reservation Multiple Access proto-\ncol) [15], and RTMAC (Real-Time Medium Access protocol) [16].\n■\nContention-based protocols with scheduling mechanism: These\nprotocols focus on packet scheduling at nodes and also schedul-\ning nodes for access to the channel. The scheduling is done\nin such a way that all nodes are treated fairly and no node\nis starved of bandwidth. These protocols can provide priori-\nties among flows whose packets are queued at nodes. Some of\nthe existing scheduling-based protocols are DWOP (Distributed\nWireless Ordering Protocol) [17], DLPS (Distributed Laxity-based\nPriority Scheduling) [18], and DPS (Distributed Priority Schedul-\ning) [19].\nContention-based protocols that use single-channel for communica-\ntion cannot completely eliminate contention for the channel. In the\ncase of WMNs the end-to-end throughput significantly reduces due\nto the accumulating effect of the contention in the multi-hop path.\nFurther, an ongoing transmission between a pair of nodes refrains\nall the nodes which are in a two-hop neighborhood of nodes partic-\nipating in the transmission from transmitting on the channel during\nthe transmission period. To overcome these problems multi-channel\nMAC and multi-channel multi-radio MAC protocols are proposed.\n■\nMulti-Channel MAC (MMAC): Multi-channel MAC [20] is a link layer\nprotocol where each node is provided with only one interface, but\nto utilize the advantage of multi-channel communication, the inter-\nface switches among different channels automatically. In MMAC the\ncommunication time is split into a number of beacon intervals. In\nthe beginning of each beacon interval, during an ATIM (Ad hoc\nTraffic Indication Message) window period all the nodes in the net-\nwork tune their radio to a common control channel and negotiate\nfor the channel to be used for the remaining period of the beacon\ninterval. Each node maintains a data structure called PCL (Preferred\nChannel List — usage of the channels within the transmission range\nof the node). When a source node S1 wants to send data to re-\nceiver node R1, during the ATIM window node S1 sends an ATIM\npacket with its PCL. Upon receiving the ATIM packet from node\nS1, node R1 compares the PCL of node S1 with its PCL and de-\ncides which channel is to be used during the beacon interval. Then\nnode R1 sends an ATIM-ACK carrying the ID of the preferred chan-\nnel. Node S1, on receiving the ATIM-ACK, confirms the reservation\nby sending an ATIM-RES packet to node R1. When other nodes in\nthe vicinity of node R1 hear the ATIM-ACK, they choose a differ-\nent channel for their communication. The throughput of MMAC is\nhigher than that of IEEE 802.11 when the network load is high. This\nincrease in throughput is due to the fact that each node uses an\n" }, { "page_number": 30, "text": "An Introduction to Wireless Mesh Networks\n■\n19\northogonal channel, thereby increasing the number of simultaneous\ntransmissions in the network. Though MMAC increases the through-\nput, there are some drawbacks with it. When a node has to send a\npacket to multiple destinations, it can send only to one destination\nin a beacon interval, because the nodes have to negotiate during\nthe ATIM window in the control channel. Due to this restriction the\nper-packet delay increases significantly. MMAC does not have any\nscheme for broadcasting.\nSlotted Seeded Channel Hopping protocol (SSCH) is another multi-\nchannel link layer protocol using a single transceiver [21]. SSCH is\nimplemented in software over an IEEE 802.11-compliant wireless\nNetwork Interface Card (NIC). SSCH uses a distributed mechanism\nfor coordinating the channel switching decision. By this channel\nhopping at each node, packets of multiple flows in the interfering\nrange of each other are transmitted simultaneously in an orthogonal\nchannel. This improves the overall capacity of the multi-hop wire-\nless network if the network traffic pattern has multiple flows in the\ninterfering range of each other. Each node in the network finds the\nchannel hopping schedule for it and schedules the packets within\neach channel. Each node transmits its channel hopping schedule to\nall its neighboring nodes and updates its channel hopping schedule\nbased on traffic pattern. SSCH yields significant capacity improve-\nment in both single-hop and multi-hop network scenarios.\n■\nMulti-Radio Multi-Channel MAC: In the application scenarios where\nthe cost of the node and power consumption are not big issues,\nnodes can be provided with multiple wireless interfaces which are\ntuned to non-overlapping channels and can communicate simultane-\nously with multiple neighboring nodes. If nodes have multiple inter-\nfaces, then the MAC protocol has to handle the orthogonal channel\nassignment to each interface and schedule the packets to the ap-\npropriate interface. The Multi-radio Unification Protocol (MUP) [22]\nis one such protocol to coordinate the operation of the multiple\nwireless NICs tuned to non-overlapping channels. MUP works as a\nvirtual MAC which requires no changes to the higher layer proto-\ncols and works with other nodes which do not have MUP. So these\ntype of nodes can be added incrementally even after deployment.\nFor the higher layer protocols the MUP looks like a single MAC run-\nning. It monitors the channel quality on each of the NICs to each of\nits neighbors. When the higher layer protocol sends packets to the\nMUP, it selects the right interface to forward the packets.\nKyasanur and Vaidya [23,24] proposed a link layer protocol for the\nscenario of nodes having more than one interface. The interfaces of\na node are grouped into two fixed interfaces where interfaces are as-\nsigned a channel for long intervals of time and switchable interfaces\n" }, { "page_number": 31, "text": "20\n■\nSecurity in Wireless Mesh Networks\nwhere interfaces are assigned dynamically for short spans of time.\nThe channel assigned to fixed interfaces is called a fixed channel and\nthat assigned to switchable interfaces is called a switchable channel.\nEach node has both a fixed channel and a switchable channel. Dur-\ning a flow initiation, each node finds the channel in the switchable\ninterface based on the fixed channel of the next-hop neighbor to\ntransmit the data to it. Once the switchable interfaces are switched\nto a channel there is no need for switching the channel for the sub-\nsequent packets for that flow unless another flow requires channel\nswitching on the switchable interface.\n1.4.4\nRouting\nThere are numerous routing protocols proposed for ad hoc networks in the\nliterature. Because WMNs are multi-hop networks, the protocols designed\nfor ad hoc networks also work well for WMNs. The main objective of those\nprotocols is quick adaptation to the change in a path when there is path\nbreak due to mobility of the nodes. Current deployments of WMNs make\nuse of routing protocols proposed for ad hoc networks such as AODV (Ad\nhoc On-Demand Distance Vector) [25], DSR (Dynamic Source Routing) [26],\nand TBRPF (Topology Broadcast based on Reverse Path Forwarding) [27].\nHowever, in WMNs the mesh routers have minimal mobility and there is no\npower constraint, whereas the clients are mobile with limited power. Such\ndifference needs to be considered in developing efficient routing protocols\nfor WMNs. As the links in the WMNs are long lived, finding a reliable and\nhigh throughput path is the main concern rather than quick adaptation to\nlink failure as in the case of ad hoc networks.\n1.4.4.1\nRouting Metrics for WMNs\nMany ad hoc routing protocols such as AODV and DSR use hop count as a\nrouting metric. This is not well suited for WMNs for the following reasons.\nThe basic idea in minimizing the hop count for a path is that it reduces the\npacket delay and maximizes the throughput. But the assumption here is that\nlinks in the path either work perfectly or do not work at all and all links are\nof equal bandwidth. A routing scheme that uses the hop count metric does\nnot take the link quality into consideration. A minimum hop count path\nhas higher average distance between nodes present in that path compared\nto a higher hop count path. This reduces the strength of the signal received\nby the nodes in that path and thereby increases the loss ratio at each link\n[28]. Hence, it is always possible that a two-hop path with good link quality\nprovides higher throughput than a one-hop path with a poor/lossy link. A\nrouting scheme that uses the hop count metric always chooses a single-\nhop path rather than a two-hop path with good link quality. The wireless\n" }, { "page_number": 32, "text": "An Introduction to Wireless Mesh Networks\n■\n21\nlinks usually have asymmetric loss rate as reported in [29]. Hence, new\nrouting metrics based on the link quality are proposed in the literature.\nThey are ETX (Expected Transmission Count), per-hop RTT (Round-Trip\nTime), and per-hop packet pair. Couto et al. proposed ETX to find a high\nthroughput path in WMNs [28]. The metric ETX is defined as the expected\nnumber of transmissions (including retransmissions) needed to successfully\ndeliver a packet over a link. As per IEEE 802.11 standard, a successful\ntransmission requires acknowledgment back to the sender. ETX considers\ntransmission loss probability in both directions, which may not be equal as\nstated earlier. All nodes in the network compute the loss probability to and\nfrom its neighbors by sending probe packets. If pf and pr are respectively\nthe loss probability in forward and reverse direction in a link, then the\nprobability that a packet transmission is not successful in a link is given by\np = 1 −(1 −pf )(1 −pr). The expected number of transmissions on that\nlink is computed as ETX =\n1\n1−p. In [30] the routing metrics based on link\nquality are compared with the hop count metric. The routing metric based\non link quality performs better than hop count if nodes are stationary. The\nhop count metric outperforms the link quality metric if nodes are mobile.\nThe main reason for this is that the ETX metric cannot quickly track the\nchanges in the value of the metric. If the nodes are mobile, the ETX value\nchanges frequently as the distance between the nodes changes.\nAs stated earlier, to improve the throughput the multi-radio multi-channel\narchitecture is used in WMNs. In this case the routing metric based on link\nquality alone is not sufficient. It should also consider the channel diversity\non the path. A new routing metric WCETT (Weighted Cumulative Expected\nTransmission Time) is proposed in [31], which takes both link quality and\nchannel diversity into account. The link quality is measured by a per-link\nmetric called ETT (Expected Transmission Time; expected time to transmit\na packet of a certain size over a link). If the size of the packet is S and\nthe bandwidth of the link is B, then ETT = ETX ∗S\nB. The channel diver-\nsity in the path is measured as follows. If X j is the sum of ETTs of the\nlinks using the channel j in the path, then channel diversity is measured\nas max1≤j≤kX j, where k is the number of orthogonal channels used. The\npath metric for path p with n links and k orthogonal channels is calculated\nas\nWCETT (p) = (1 −β) ∗\nn\n\u0002\ni=1\nETTi + β ∗max1≤j≤kX j,\nwhere β is a tunable parameter subject to 0 ≤β ≤1. WCETT can achieve\na good trade-off between delay and throughput as it considers both link\nquality and channel diversity in a single routing metric.\nThe WCETT metric considers the quality of links and the intra flow\ninterference along the path. But it fails to take into account inter flow\n" }, { "page_number": 33, "text": "22\n■\nSecurity in Wireless Mesh Networks\ninterference on the path. In [32], a new routing metric MIC (Metric of In-\nterference and Channel switching) is proposed for multi-channel multi-\nradio WMNs. This new metric considers the quality of links, inter flow\ninterference, and intra flow interference altogether. This metric is based\non Interference-Aware Resource Usage (IRU) and Channel Switching Cost\n(CSC) metrics to find the MIC for a given path. IRU captures the differences\nin the transmission rate and the loss ratios of the wireless link and the\ninter flow interference. The IRU metric for a link k which uses channel c\nis calculated as IRU k(c) = ETT k(c) ∗Nk(c), where ETT k(c) is the expected\ntransmission time of the link k on the channel c, and Nk(c) is the number\nof nodes interfering with the transmission of the link k on channel c. The\nCSC metric captures the intra flow interference along the path. CSC for a\nnode i is assigned a weight w1 if links in the path connected to it have\ndifferent channels assigned, and w2 if they are the same, 0 ≤w1 < w2. The\npath metric for a given path p, MIC(p), is calculated as follows:\nMIC(p) = α ∗\n\u0002\n(link l ε p)\nIRU l +\n\u0002\n(node i ε p)\nCSCi.\nHere α is a positive factor which gives a trade-off between benefits of IRU\nand CSC.\n1.4.4.2\nRouting Protocols for WMNs\nIn [30], the authors proposed an LQSR (Link Quality Source Routing) pro-\ntocol. It is based on DSR and uses ETX as the routing metric. The main\ndifference between LQSR and DSR is getting the ETX metric of each link\nto find out the path. During the route discovery phase, the source node\nsends a Route Request (RREQ) packet to neighboring nodes. When a node\nreceives the RREQ packet, it appends its own address to the source route\nand the ETX value of the link in which the packet was received. The des-\ntination sends the Route Reply (RREP) packet with a complete list of links\nalong with the ETX value of those links. Because the link quality varies\nwith time, LQSR also propagates the ETX value of the links during the data\ntransmission phase. On receiving a data packet, an intermediate node in\nthe path updates the source route with the ETX value of the outgoing link.\nUpon receiving the packet, the destination node sends an explicit RREP\npacket back to the source to update the ETX value of links in the path.\nLQSR also uses a proactive mechanism to update the ETX metric of all\nlinks by piggybacking Link-Info messages to RREQ messages occasionally.\nThis Link-Info message contains the ETX value of all the links incident on\nthe originating node.\nA new routing protocol for multi-radio multi-channel WMNs called\nMulti-Radio Link Quality Source Routing (MR-LQSR) is proposed in [31],\nwhich uses WCETT as a routing metric. The neighbor node discovery and\n" }, { "page_number": 34, "text": "An Introduction to Wireless Mesh Networks\n■\n23\npropagating the link metric to other nodes in the network in MR-LQSR are\nthe same as that in the DSR protocol. But assigning the link weight and\nfinding the path weight using the link weight are different from DSR. DSR\nuses equal weight to all links in the network and implements the shortest\npath routing. But MR-LQSR uses a WCETT path metric to find the best path\nto the destination.\nIn [32], the authors showed that, if a WCETT routing metric is used\nin a link state routing protocol, it is not satisfying the isotonicity property\nof the routing protocol and leads to formation of routing loops. To avoid\nthe formation of routing loops by the routing metrics, they proposed Load\nand Interference Balanced Routing Algorithm (LIBRA) [32], which uses MIC\nas the routing metric. In LIBRA a virtual network is formed from the real\nnetwork and decomposed the MIC metric into isotonicity link weight as-\nsignment on the virtual network. The objective of MIC decomposition is\nto ensure that LIBRA can use efficient algorithms such as Bellman–Ford or\nDijkstra’s algorithm to find the minimum weight path on the real network\nwithout any forwarding loops.\n1.4.5\nTransport Layer\nThere are several reliable transport protocols proposed for ad hoc networks.\nSome of them are modified versions of TCP (Transmission Control Protocol)\nthat work well in ad hoc networks and others are designed specifically for\nan ad hoc network scenario from scratch.\nTCP is the de facto standard for end-to-end reliable transmission of data\non the Internet. TCP was designed to run efficiently on wireline networks.\nUsing the TCP protocol on a wireless network degrades the performance\nof the network in terms of reduction in throughput and unfairness to the\nconnections. This degradation in performance is due to the following rea-\nsons. The Bit Error Rate (BER) in wireless networks is very high compared\nto wireline networks. Frequency of path break in wireless networks is high\ndue to mobility of nodes in ad hoc networks. If the packets get dropped in\nthe network due to these reasons, the TCP sender misinterprets this event\nas congestion and triggers the congestion control mechanism to reduce\nthe congestion window size. This reduces the effective throughput of the\nnetwork.\n■\nTCP Variants for Wireless Networks: To solve the problem of degra-\ndation of throughput of TCP over wireless networks, various modifi-\ncations to TCP protocols have been proposed. These modifications\nare mainly based on differentiating the congestion loss and non-\ncongestion loss at the TCP sender when there is a packet loss in\nthe network. The proposed protocols [33] and [34] rely on coop-\neration from the network, i.e., the intermediate nodes inform the\n" }, { "page_number": 35, "text": "24\n■\nSecurity in Wireless Mesh Networks\nsource regarding the status of a path. In ELFN (Explicit Link Failure\nNotification) [33], the intermediate node informs the sender about\nthe link failure explicitly. When the sender is informed that the link\nhas failed, it disables its retransmission timer and enters into standby\nmode. In the standby mode the sender probes the network to check\nif the network connection is re-established by sending a packet from\nthe congestion window periodically. Upon receiving an ACK from\nthe receiver, i.e., after the connection is established, the sender re-\nsumes its normal operation. In TCPF (TCP-Feedback) [34], when an\nintermediate node detects path break, it sends an RFN (Route Fail-\nure Notification) message to the TCP sender. On receiving an RFN\nmessage, the TCP sender goes to snooze state. In this state the TCP\nsender stops sending packets and freezes all its variables such as\nretransmission buffer, congestion window, and packet buffer. Once\nthe route is established again, the intermediate node sends an RRN\n(Route Re-established Notification) message to the sender. Upon\nreceiving an RRN message from an intermediate node, the sender\nresumes its transmission using the same variable values that were\nbeing used prior to interruption. To avoid an infinite wait for an RRN\nmessage, TCPF uses a route failure timer, which is the worst-case\nroute re-establishment time.\n■\nOther Transport Protocols for Wireless Networks: In [35], a transport\nprotocol for wireless networks was proposed by not modifying the\nexisting TCP protocol. This is done by introducing a thin layer called\nATCP between the network layer and transport layer and it is invis-\nible to transport layer. This makes nodes with ATCP and without\nATCP interoperable with each other. ATCP gets information about\ncongestion in the network from the intermediate nodes through ECN\n(Explicit Congestion Notification) and ICMP messages. Based on this,\nthe source node distinguishes congestion and non-congestion losses\nand takes the appropriate action.\n■\nWhen the TCP sender identifies any network partitioning, it goes\ninto persist state and stops all the outgoing transmissions.\n■\nWhen the TCP sender notices any loss of packets in the network\ndue to channel error, it retransmits the packet without invoking\nany congestion control.\n■\nWhen the network is truly congested, it invokes the TCP con-\ngestion control mechanism.\n1.4.6\nGateway Load Balancing\nIn WMNs the gateway nodes are connected to the backhaul network, i.e.,\nto the Internet, which provides Internet connectivity to all nodes in the net-\nwork. So the gateway may become a bottleneck for the connections to the\n" }, { "page_number": 36, "text": "An Introduction to Wireless Mesh Networks\n■\n25\nInternet. As many clients in the network generate traffic to the gateway, the\navailable bandwidth should be utilized effectively. The traffic generated by\nclient nodes aggregates at gateway nodes in the WMN. If some of the gate-\nway nodes are highly loaded and other gateway nodes are lightly loaded,\nit creates load imbalance between gateway nodes, which leads to packet\nloss and results in a degradation in network performance. Hence, load bal-\nancing across gateway nodes in WMNs improves bandwidth utilization and\nalso increases network throughput.\nLoad balancing across gateway nodes is obtained by distributing the\ntraffic generated by the network to the backhaul network through all gate-\nway nodes in the WMNs. The load balancing across multiple gateway nodes\ncan be measured quantitatively by a metric called Index of Load Balance\n(ILB) [36] which is calculated as follows.\nLoad index (LI) of a gateway i is defined as the fraction of the gateway’s\nbackhaul link utilized by a given node k, L I (i) =\n\u0003\nk∈N βk(i)∗Tk\nC(i)\n, where βk(i)\nis the fraction of node k’s traffic that is sent through gateway i, Tk is the\ntotal traffic generated by node k, and C(i) is the capacity of the backhaul\nlink connected to the gateway node i. The LI value ranges from 0 to 1,\nwith 1 representing 100 percent loaded gateway. The ILB of the network\nis calculated as\nILB = max{LI(i)} −min{LI(i)}\nmax{LI(i)}\nTherefore a perfectly balanced network has ILB equal to zero and a highly\nimbalanced network has ILB equal to one. The objective of all load bal-\nancing techniques is to obtain ILB values as small as possible. Several tech-\nniques for load balancing across gateways were proposed in the literature.\nSome of them are discussed in this section.\n■\nMoving Boundary-Based Load Balancing: A flexible boundary is de-\nfined for each gateway and the nodes which fall in the boundary are\ndirected to communicate through that gateway. To adopt to varia-\ntions in the traffic, the region of boundary is periodically redefined.\nThe boundary can be defined in two different ways: (1) in a shortest\npath-based moving boundary approach, the boundary region for a\ngateway node is defined by distance of the node from the gateway,\nand (2) in a load index-based moving boundary approach, the gate-\nways announce their load Index and the nodes join lightly loaded\ngateways. In this scheme the lightly loaded gateway serves more\nnodes and the heavily loaded gateway serves fewer nodes.\n■\nPartitioned Host-Based Load Balancing: Here, the nodes in the net-\nwork are grouped, and each group is assigned to a particular gate-\nway. The main difference compared to the moving boundary-based\n" }, { "page_number": 37, "text": "26\n■\nSecurity in Wireless Mesh Networks\nload balancing is that no clear boundary is defined. This can be\ndone in both a centralized and distributed way. In the centralized\nmethod, a central server assumes the responsibility of assigning the\ngateway to the nodes. The central server collects the complete infor-\nmation about the gateway nodes and traffic requirements of all the\nnodes and then allocates nodes to the gateways. In the distributed\nmethod, a logical network is formed by the gateway nodes. Each\nnode is associated with a gateway node known as a dominating gate-\nway through which traffic generated by this node reaches the Inter-\nnet. The nodes in the network periodically update their dominating\ngateway about their traffic demand. The gateway nodes exchange\ninformation about their load and capacity information through the\nlogical network. When a gateway is highly loaded, hand-over takes\nplace, i.e., the gateway delegates some nodes to other gateways\nwhich are lightly loaded.\n■\nProbabilistic Stripping-Based Load Balancing: In the techniques dis-\ncussed above, each node in the network utilizes only one gateway,\nwhich may not lead to perfect load balancing among the gateways.\nIn a probabilistic stripping-based load balancing scheme, each node\nutilizes multiple gateways simultaneously, which gives perfect load\nbalancing theoretically. In this technique each node identifies all the\ngateway nodes in the network and attempts to send a fraction of its\ntraffic through every gateway. Hence, the total traffic is split among\nmultiple gateways. This technique is applicable in the case where\nthe load can be split for sending through multiple gateways.\n1.4.7\nSecurity\nAs mentioned earlier, due to the unique characteristics of WMNs, they are\nhighly vulnerable to security attacks compared to wired networks. Design-\ning a foolproof security mechanism for WMNs is a challenging task. The\nsecurity can be provided in various layers of the protocol stack. Current\nsecurity approaches may be effective against a particular attack in a spe-\ncific protocol layer, but they lack a comprehensive mechanism to prevent\nor counter attacks in different protocol layers. The following issues pose\ndifficulty in providing security in WMNs:\n■\nShared Broadcast Radio Channel: In a wired network, a dedicated\ntransmission line is provided between the nodes. But the wireless\nlinks between the nodes in WMNs are broadcast in nature, i.e., when\na node transmits, all the nodes within its direct transmission range\nreceive the data. Hence, a malicious node could easily obtain data\nbeing transmitted in the network if it is placed in the transmission\n" }, { "page_number": 38, "text": "An Introduction to Wireless Mesh Networks\n■\n27\nrange of mesh routers or a mesh client. For example, if you have a\nWMN and so does your neighbor, then there is a scope for either\nsnooping into private data or simply hogging the available band-\nwidth of a neighboring, but alien node.\n■\nLack of Association: In WMNs, the mesh routers form a fixed mesh\ntopology which forms a backbone network for the mobile clients.\nHence, the clients can join or leave the network at any time through\nthe mesh routers. If no proper authentication mechanism is provided\nfor association of nodes with WMNs, an intruder would be able to\njoin the network quite easily and carry out attacks.\n■\nPhysical Vulnerability: Depending on the application of WMNs, the\nmesh routers are placed on lampposts and rooftops, which are vul-\nnerable to theft and physical damage.\n■\nLimited Resource Availability: Normally, the mesh clients are limited\nin resources such as bandwidth, battery power, and computational\npower. Hence, it is difficult to implement complex cryptography-\nbased mechanisms at the client nodes. As mesh routers are resource\nrich in terms of battery power and computational power, security\nmechanisms can be implemented at mesh routers. Due to wireless\nconnectivity between mesh routers, they also have bandwidth con-\nstraints. Hence, the communication overhead incurred by the secu-\nrity mechanism should be minimal.\n1.4.8\nPower Management\nThe energy efficiency of a node in the network is defined as the ratio of\nthe amount of data delivered by the node to the total energy expended.\nHigher energy efficiency implies that a greater number of packets can be\ntransmitted by the node with a given amount of energy resource. The main\nreasons for power management in WMNs are listed below.\n■\nPower Limited Clients: In WMNs, though the mesh routers do not\nhave limitations on power, clients such as PDAs and IP phones\nhave limited power as they are operated on batteries. In the case of\nHybrid WMNs, clients of the other networks that are connected to\nthem, such as sensor networks, can be power limited. Hence, power\nefficiency is of major concern in WMNs.\n■\nSelection of Optimal Transmission Power: In multi-hop wireless net-\nworks, the transmission power level of wireless nodes affects con-\nnectivity, interference, spectrum spatial reuse, and topology of the\nnetwork. Reducing the transmission power level decreases the in-\nterference and increases the spectrum spatial reuse efficiency and\nthe number of hidden terminals. An optimal value for transmission\n" }, { "page_number": 39, "text": "28\n■\nSecurity in Wireless Mesh Networks\npower decreases the interference among nodes, which in turn in-\ncreases the number of simultaneous transmissions in the network.\n■\nChannel Utilization: In multi-channel WMNs, the reduction in trans-\nmission power increases the channel reuse, which increases the\nnumber of simultaneous transmissions that improves the overall ca-\npacity of the network. Power control becomes very important for\nCDMA-based systems in which the available bandwidth is shared\nby all the users. Hence, power control is essential to maintain the\nrequired signal-to-interference ratio (SIR) at the receiver and to\nincrease the channel reusability.\nSeveral power efficient MAC protocols and power-aware routing proto-\ncols are proposed for ad hoc networks to efficiently utilize limited energy\nresource available in mobile nodes. These protocols consider all the nodes\nin the network power limited. In WMNs, some nodes are power limited\nand others have no limitation on power. So, when a power-efficient pro-\ntocol is used in WMNs, it would not utilize the resource-rich mesh routers\nto reduce power consumption on power-limited mesh clients. Hence, new\nprotocols are required which consider both types of nodes and efficiently\nutilize the power of the client nodes.\n1.4.9\nMobility Management\nIn WMNs the mobile clients get network access by connecting to one of\nthe mesh routers in the network. When a mobile client moves around the\nnetwork, it switches its connectivity from one mesh router to another. This is\ncalled hand-off or hand-over. In WMNs the clients should have capability to\ntransfer connectivity from one mesh router to another to implement hand-\noff technique efficiently. Some of the issues in handling hand-offs in WMNs\nare discussed below.\n■\nOptimal Mesh Router Selection: Each mesh client connects to one of\nthe mesh routers in the WMN. Normally, each mesh client chooses\nthe mesh router based on the signal strength it receives from the\nmesh routers. When a mobile client is in the transmission range of\nmultiple mesh routers, it is very difficult to clearly decide to which\nmesh router the mobile client must be assigned.\n■\nDetection of Hand-off: Hand-off may be client initiated or network\ninitiated. In the case of client initiated, the client monitors the signal\nstrength received from the current mesh router and requests a hand-\noff when the signal strength drops below a threshold. In the case of\nnetwork initiated, the mesh router forces a hand-off if the signal from\n" }, { "page_number": 40, "text": "An Introduction to Wireless Mesh Networks\n■\n29\nthe client weakens. Here the mesh router requires information from\nother mesh routers about the signal strength they receive from the\nparticular client and deduces to which mesh router the connection\nshould be handed over.\n■\nHand-Off Delay: During hand-off, the existing connections between\nclients and network get interrupted. Though the hand-off gives con-\ntinuous connectivity to the roaming clients, the period of interrup-\ntion may be several seconds. All ongoing transmissions of the client\nare transferred from the current mesh router to a new mesh router.\nThe time taken for this transfer is called hand-off delay. The delay of\na few seconds may be acceptable for applications like file transfer,\nbut for applications that require real-time transport such as interac-\ntive VoIP (Voice-over-IP) or videoconferencing, it is unacceptable.\n■\nQuality of Hand-Off: During hand-off some number of packets may\nbe dropped due to hand-off delay or interruption on the ongoing\ntransmission. The quality can be measured by the number of packets\nlost per hand-off. A good quality hand-off provides a low packet loss\nper hand-off. The acceptable amount of packet loss per hand-off\ndiffers between applications.\nThe hand-off mechanisms in cellular networks are studied in [37] and\n[38]. When a user moves from the coverage area of one BS to the adjacent\none, it finds an uplink–downlink channel pair from the new cell and drops\nthe link from the current BS. In WLANs, whenever a client moves from one\nAP to another, the link has to be reconfigured manually. In this case, all\nongoing connections are terminated abruptly. It may be applicable in LAN\nenvironments as the clients have limited mobility around a limited area. But\nin the case of WMNs, the mesh clients may constantly roam around different\nmesh routers. Here, manual reconfiguration of mesh clients, whenever the\nclient moves from one mesh router to another, is a difficult task. So the\nhand-off has to be done automatically and transparently. The users should\nnot feel that the existing connections are transferred from one mesh router\nto another. For applications such as VoIP and IPTV in WMNs, sophisticated\nand transparent hand-off techniques are required.\n1.4.10\nAdaptive Support for Mesh Routers and Mesh Clients\nCompared to other networking technologies where all the nodes in the net-\nwork are considered to have similar characteristics, WMNs have different\ncharacteristics between mesh routers and mesh clients. The main differ-\nences between them which make the need for new networking protocols\nfor WMNs are\n" }, { "page_number": 41, "text": "30\n■\nSecurity in Wireless Mesh Networks\n■\nMobility: In many applications of WMNs, the mesh routers form a\nfixed backbone network by placing the mesh routers at fixed loca-\ntions such as rooftops and lampposts. So the mesh routers are con-\nsidered immobile, but the clients in the mesh network are highly\nmobile and can be connected to any mesh router based on signal\nstrength received from different mesh routers.\n■\nResource Availability: Normally, mesh routers are operated with\nelectric power rather than battery power. They are placed in loca-\ntions where the powerline is available, so the mesh routers do not\nhave energy constraints. But the clients are operating with battery\npower and are considered energy constrained.\nThe existing protocols for ad hoc networks consider the characteristics of\nall nodes in the same way. The energy-aware protocols consider all nodes\nin the network battery operated. The protocols that take into account the\nmobility of nodes in the network consider all nodes in the network mobile.\nFor example, a routing protocol designed for networks with high mobility\nand limited power when used in WMNs does not utilize the limited mobility\nand rich energy resource nature of mesh routers. Hence, it fails to improve\nthe performance of WMNs. But due to the characteristics of mesh routers,\nthe routing protocols become simple and efficient. So WMNs need efficient\nprotocols that consider the differences between the mesh routers and mesh\nclients to improve the performance of WMNs.\n1.4.11\nIntegration with Other Network Technologies\nThe integration of WMNs with other existing network technologies such\nas cellular, WiFi, WiMAX, WiMedia, and sensor networks can be achieved\nby bridging functions at the mesh routers. These bridging functions can be\nprovided by adding network interfaces corresponding to the networking\ntechnology that the mesh router has to support. There are several issues to\nbe addressed in integrating multiple networking technologies with WMNs:\n■\nComplexity of Mesh Router: The integration of multiple networking\ntechnologies with the mesh network increases the complexity of the\nmesh routers. For each networking technology to be supported by a\nmesh router, a network interface should be provided. This increases\nthe hardware and software complexity of the mesh routers.\n■\nCost of Mesh Router: The networking hardware or network inter-\nface for different networking technologies are not the same. Each\nnetworking technology needs specially designed hardware to oper-\nate on. Mesh routers have to be provided with the same number of\ninterfaces as the number of networking technologies supported by\nthem. This increases the cost of mesh routers.\n" }, { "page_number": 42, "text": "An Introduction to Wireless Mesh Networks\n■\n31\n■\nServices Provided by Integrated WMNs: The services provided by\ndifferent networking technologies are different. Services not pro-\nvided by IEEE 802.11 can be provided by cellular networks. Simi-\nlarly the services provided by sensor networks cannot be provided\nby cellular networks. The integration of other networking technolo-\ngies with WMNs provides many services to the users that are not\nprovided by WMNs alone. Depending on the service requirement,\nthe required networking technologies can be integrated with WMNs.\n■\nInter-Operability of Network Technologies: The protocols for differ-\nent network technologies are independent and operating them to-\ngether is a difficult task. For example, the routing protocols used by a\ncellular network and an IEEE 802.11 network are not the same. Fur-\nther, the MAC protocols used by different networking technologies\nare not inter-operable. So the inter-operability of different network-\ning technologies necessitates new software architectures or middle-\nware implementations over the mesh networking platform.\nThough the integration of multiple networking technologies with WMNs\nis a difficult job, the services rendered by this necessitate the researchers\nto come up with a feasible solution. The development of new network\narchitectures and middleware solutions may solve some of these problems.\nThe problem of implementation of many network interfaces in a single\nmesh router can be solved by using software-defined radios. The software-\ndefined radio system is a software-based communication system for mod-\nulation and demodulation of radio signal. This is done by advanced signal\nprocessing techniques implemented in a digital computer or in a reconfig-\nurable digital electronic system. This technique produces different radios\nthat can receive and transmit a new form of radio protocol just by run-\nning different software rather than designing new hardware. This helps in\nreducing the number of networking interfaces in mesh routers.\n1.4.12\nDeployment Considerations\n■\nScenario of Deployment: The capability required for deployments of\ndifferent WMNs is not the same. For example, WMN deployment for\ncommunity networking to share network resources among people\nis not the same as for rescue operations. Some of the deployment\nscenarios in which the deployment issues vary are\n■\nEmergency Operation Deployment:\nThis kind of application\nscenario demands a quick deployment of a communication\nbackbone network through which the mobile devices can com-\nmunicate. For example, during disasters like flood, fire, and\nearthquake all the existing communication network infra-\nstructure might be destroyed. Hence, a quick deployment of\n" }, { "page_number": 43, "text": "32\n■\nSecurity in Wireless Mesh Networks\na backbone communication network is essential. Most impor-\ntantly, the network should provide support for time-sensitive\ntraffic such as voice and video. The network should also provide\nsupport for different networking technologies to communicate\nusing this network. Hence, the mesh routers should provide in-\nterfaces for other existing technologies which allow people to\ncommunicate using any communication equipment they have.\n■\nCommercial Broadband Access Deployment:\nThe aim of this\ndeployment is to provide an alternate network infrastructure for\nwireless communications in urban areas and areas where a tradi-\ntional cellular BS cannot handle the traffic volume. This scenario\nassumes significance as it provides very low cost per bit trans-\nferred compared to the cellular network infrastructure. Another\nmajor advantage of this application is the resilience to failure of\na certain number of nodes. Addressing, configuration, position-\ning of relaying nodes, redundancy of nodes, and power sources\nare the major issues in deployment. Billing, provisioning of QoS,\nsecurity, and handling mobility are major issues that the service\nprovider needs to address.\n■\nHome Network Deployment: The deployment of a home area\nnetwork needs to consider the limited range of the devices that\nare to be connected by the network. Given the short transmis-\nsion ranges of a few meters, it is essential to avoid network\npartitions. Positioning of mesh routers at certain key locations\nof a home area network can solve this problem; also network\ntopology should be decided so that every mesh router is con-\nnected through multiple neighbors for availability.\n■\nCost of Deployment: The commercial deployment of a communi-\ncations infrastructure using a WMN essentially eliminates the re-\nquirement of laying cables and maintaining them. Hence, the cost\nof deployment is much less than that of the wired infrastructure.\nOnly the mesh routers have to be placed in appropriate locations\nfor efficient coverage. The mesh router manufacturers are providing\nmesh routers for outdoor placements. Mesh routers can be placed on\npoles on the street, which reduces the cost of deployment of mesh\nnetworks.\n■\nIncremental Deployment: In any WMN deployment, the coverage of\na geographical area can be extended by adding mesh routers incre-\nmentally. With minimum configuration, the network starts function-\ning and mesh routers can be added incrementally for expanding the\nsize of the network. For example, during the community networking\ndeployment process whenever a mesh router is installed, it can be\ncommissioned.\n" }, { "page_number": 44, "text": "An Introduction to Wireless Mesh Networks\n■\n33\n■\nShort Deployment Time: Compared to any wired communication\ninfrastructure, WMNs have less deployment time due to absence of\nlaying cables. Wiring the dense urban region is extremely difficult\nand time consuming, in addition to the inconvenience caused. Mesh\nrouters can be placed based on the area of coverage and number\nof active users in the area. They can be deployed even on rooftops,\nprovided that electrical power is available.\n■\nAuto-Configurability: The incremental deployment of mesh networks\nto increase the coverage area or number of users leads to changes in\ntopology of the network at later stages. The lossy nature of the wire-\nless medium changes due to environmental changes, which leads\nthe routing protocols to change the path very often. Due to this, the\nnetwork needs re-configuration very often.\n■\nOperational Integration with Other Infrastructure: Operational inte-\ngration with other networking technologies such as satellite, cellular,\nand sensor networks can be considered to improve the performance\nor provide additional services to the end users. In the commercial\nworld, the WMNs that service a given urban region can interoperate\nwith the cellular infrastructure to provide better QoS and smooth\nhand-offs across the networks. Hand-offs to a different network can\nbe done to avoid call drops when a mobile node with an active call\nmoves into a region where service is not provided by the current\nnetwork.\n■\nArea of Coverage: In most of the cases, the area of coverage of\nWMNs is determined by the nature of application for which the net-\nwork is set up. For example, for home networks the coverage of\nthe mesh routers is within the home or within the room in which\nthe router is placed. But in the case of wireless service providers,\nmesh routers should be covering a number of homes on a street.\nLong-range communication by fixed mesh routers can be achieved\nby means of directional antennas. The mesh routers’ and mobile\nclients’ capabilities such as transmission range and associated hard-\nware, software, and power source should match the area of coverage\nrequired.\n■\nService Availability: Service availability is defined as the ability of\na network to provide service even with failure of certain nodes. In\nWMNs the mesh routers form a fixed mesh backbone to provide\nmultiple services to the mobile clients. These mesh routers may be\nplaced in outdoor areas such as lampposts and rooftops. They are\nsubject to failure due to power failure, environmental damage, phys-\nical damage, or theft. Due to this, the services provided by a WMN\nto mobile clients may not be available in certain areas. Hence, the\nmesh routers need to be placed in such a way that failure of some\n" }, { "page_number": 45, "text": "34\n■\nSecurity in Wireless Mesh Networks\nof them does not lead to lack of service in that area. In such cases,\nredundant inactive mesh routers can be placed in such a way that,\nin the event of failure of active mesh routers, the redundant mesh\nrouters can take over their responsibilities.\n■\nChoice of Protocols: The choice of protocols at different layers of\nthe protocol stack is to be done by taking into consideration the de-\nployment scenario. The MAC protocol should ensure provisioning\nof security at link level for military applications. The routing pro-\ntocol also should be selected with care. In the case of integration\nof different networking technologies, end-to-end paths may have\ndifferent types of nodes with different capabilities. It requires rout-\ning protocols that consider the resource limitations of the nodes.\nAt the transport layer, depending upon the environment in which\nthe WMN is deployed, the connection-oriented or connectionless\nprotocols should be chosen. If the clients connected to the WMN\nare highly mobile, a frequent hand-off of the clients with the mesh\nrouters takes place. This causes the higher-layer protocols to take\nnecessary action appropriately; also, packet loss arising due to con-\ngestion, channel error, link break, and network partition is to be\nhandled differently in different applications. The timer values at dif-\nferent layers of the protocol stack should be adapted to the deploy-\nment scenario.\n1.5\nWMN Deployments/Testbeds\nFor the deployment of WMNs to be viable, they must be easy to install. This\nis particularly important for home applications where people are unwill-\ning to install highly technical networks. A number of IEEE standards such\nas 802.11, 802.15, 802.16, and 802.20 have emerged recently for wireless\nnetworks. Many task groups have been working on standardization of the\nprotocols for WMNs, which leads to the development and interoperability\nof mesh networking products from different vendors. Many testbeds have\nbeen established to carry out research and development work in WMNs.\n1.5.1\nIEEE 802.11 WMNs\nIEEE 802.11 [4] is the most popular WLAN standard that defines the spec-\nifications for the physical and MAC layer and has been adopted by many\nvendors of WLAN products. A later version of this standard is the IEEE\n802.11b [6], commercially known as WiFi. The original standards for IEEE\n802.11 promised a data rate of 1 to 2 Mbps in the license-free 2.4 GHz ISM\n(Industrial, Scientific, Medical) band. IEEE 802.11b defines operation in the\n" }, { "page_number": 46, "text": "An Introduction to Wireless Mesh Networks\n■\n35\n2.4 GHz ISM band at data rates of 5.5 and 11 Mbps. IEEE 802.11a [5] operates\nin the 5 GHz band (unlicensed national information infrastructure band).\nIt supports data rates up to 54 Mbps. IEEE 802.11e deals with the require-\nments of time-sensitive applications such as voice and video. IEEE 802.11g\naims at providing the high data rate of IEEE 802.11a in the ISM band. Under\nthe 802.11 standard, mobile clients can operate in infrastructure mode and\nad hoc mode. In infrastructure mode a mobile client communicates with\nothers through one or more APs. In ad hoc mode mobile clients can com-\nmunicate directly with each other without using an AP. The set of mobile\nclients associated with a given AP is called a Basic Service Set (BSS). A BSS\nis the basic building block of the network. BSSs are connected by means of\na Distribution System (DS) to form an extended network. Any logical point\nthrough which non-IEEE 802.11 packets enter the system is called a portal.\nPortals are also used for integrating wireless networks with the existing\nwired network. The BSS, DS, and portals together with the mobile clients\nthey connect constitute the Extended Service Set (ESS). Another working\ngroup in IEEE 802.11 [3], called 802.11s, has been formed recently to stan-\ndardize the ESS for mesh networking. It defines architecture and protocols\nbased on IEEE 802.11 MAC to create an 802.11-based Wireless Distribution\nSystem (WDS). This WDS supports both broadcast, multicast, and unicast\ndelivery using radio-aware metrics over self-configuring multi-hop topolo-\ngies. There are two main proposals for 802.11s by SEEMesh and Wi-Mesh.\nThe main features of these proposals are as follows:\n■\nSupports single and multiple radios.\n■\nWith authentication and key management procedures, it provides\nsecure key distribution and secure exchange of routing information,\nsupporting centralized and distributed models.\n■\nSupports QoS and power-efficiency-aware routing with a WDS four-\naddressing extension that supports dynamic auto-configuration of\nMAC-layer data delivery.\n■\nEnables multiple routing algorithms for MAC address-based forward-\ning with a simple Hello message for mesh discovery and association\nand supporting extended mesh discovery.\n1.5.2\nIEEE 802.15 WMNs\nThe 802.15 WPAN Task Group [39] focuses on the development of consen-\nsus standards for Personal Area Networks or short-distance wireless net-\nworks. These WPANs address wireless networking of portable and mobile\ncomputing devices such as PCs, PDAs, peripherals, cell phones, pagers,\nand consumer electronics and allow these devices to communicate and\ninteroperate.\n" }, { "page_number": 47, "text": "36\n■\nSecurity in Wireless Mesh Networks\nThe IEEE 802.15 Task Group 5 is chartered to determine the mechanisms\nthat must be present in the PHY and MAC layers of WPANs to enable\nmesh networking. A mesh network is a PAN that employs one of the two\nconnection arrangements, full mesh topology or partial mesh topology. In\nthe full mesh topology, all nodes are in the transmission range of one\nanother, i.e., each node can communicate with other nodes in one hop. In\npartial mesh topology, nodes in the network have one-hop communication\nwith a few nodes only. The 802.15 mesh networks have the following\ncapabilities:\n■\nExtension of network coverage without increasing transmit power\nor receiver sensitivity\n■\nEnhanced reliability via route redundancy\n■\nEasier network configuration\n■\nBetter battery life of device due to fewer retransmissions\n1.5.3\nIEEE 802.16 WMNs\nThe Worldwide Interoperability for Microwave Access (WiMAX) forum de-\nscribes WiMAX as “a standards-based technology enabling the delivery of\nlast mile wireless broadband access as an alternative to cable and DSL.” The\n802.16 [40] standard requires line-of-sight towers and operates in the 10 to\n66 GHz frequency band. But the 802.16a [41] extension does not require\nline-of-sight and operates in the 2 to 11 GHz frequency band. To allow the\nconsumers to connect to the Internet while moving at vehicular speeds,\nthe 802.16e [42] extension was developed. The main advantage of 802.16-\nbased mesh networks compared to 802.11 is higher coverage range and\nbandwidth. As 802.16 uses TDMA-based scheduling of channel access, it\nprovides efficient resource utilization. These advantages make 802.16 best\nsuited for WMNs. The recent draft on 802.16 [43] integrated the mesh mode\nspecification into the standard. This mesh mode supports Time Division\nDuplex (TDD), which separates downlink and uplink in time. The MAC\nframe has two sub-frames called control sub-frame and data sub-frame. Ev-\nery control sub-frame consists of 16 transmission opportunities and each\ntransmission opportunity equals seven OFDM symbols. The data sub-frame\nconsists of mini slots, which are basic units for resource allocation. The\nscheduling algorithm in 802.16 allocates the time slots in the data frame.\nThis is done by control message exchange in the control sub-frame so that\nthere is no contention in the data sub-frame. In a transmission opportunity\neach node contends for channel and runs an election algorithm to compute\nwhether or not it can win a slot, because other nodes may also try to trans-\nmit in the selected time slot. If it wins in the election algorithm, the node\nbroadcasts its schedule to all the neighbors and repeats the procedures in\n" }, { "page_number": 48, "text": "An Introduction to Wireless Mesh Networks\n■\n37\nthe next transmission time. If it fails, the node selects the next transmis-\nsion slot and continues contention until it wins. For a connection setup, a\nrequest/grant/confirm three-way handshake procedure is used.\n1.5.4\nAcademic Research Testbeds\nMany academic research institutes established testbeds to study realistic\nbehavior of WMNs. Some of them are discussed in this section.\n■\nMIT Roofnet [44–46]: MIT Roofnet is an 802.11b multi-hop network\ndesigned to provide broadband Internet connectivity to users in\napartments of Cambridge, MA. It has about 50 nodes connected\nthrough 802.11b interfaces in multi-hop fashion and connected to\nthe Internet through an Ethernet interface available in the apart-\nments. Research on Roofnet includes link-level measurements of\n802.11 interfaces, finding high-throughput routes in the face of lossy\nlinks, adaptive bit-rate selection, and developing new protocols\nwhich take advantage of radio’s unique properties. The main feature\nof Roofnet is that it is an unplanned network, i.e., no configuration\nor planning is required.\n■\nCalRadio-I [47]: California Institute for Telecommunications and\nInformation Technology developed CalRadio-I, which is a radio/\nnetworking test platform for wireless research and development.\nThis is a single integrated, wireless networking test platform which\nprovides a simple, low-cost platform development from the MAC\nlayer to a higher layer. All the MAC functionalities are coded in C\nlanguage that runs on the DSP processor. Any modification to the\nMAC protocol can be done and tested in it. CalRadio-I functions as\na test instrument, an AP, and as a WiFi client.\n■\nBWN-Mesh Testbed at Georgia Tech [48]: The WMN tested by the\nBroadband and Wireless Network (BWN) Lab at Georgia Institute of\nTechnology consists of 15 IEEE 802.11b/g-based mesh routers. Using\nthis mesh network testbed, various experiments to investigate the ef-\nfects of inter-router distance, backhaul placement, and clustering are\nperformed by varying the mobility of the nodes. Other testbeds in\nthe lab such as next-generation Internet testbed as backhaul access\nto the Internet are connected to a mesh testbed. The measurements\nusing this testbed reveal that existing protocols for wireless ad hoc\nnetworks such as TCP for transport layer, AODV for network layer,\nand IEEE 802.11g for MAC do not perform well in terms of end-\nto-end delay and throughput in WMNs. So the research at BWN is\nfocused on adaptive protocols for transport, routing, and MAC layers\nand their cross-layer design. Integration of other network technol-\nogy testbeds such as WSNs (Wireless Sensor Networks), WSANs\n" }, { "page_number": 49, "text": "38\n■\nSecurity in Wireless Mesh Networks\n(Wireless Sensor and Actor Networks), next-generation Internet, and\nWiMAX with WMNs testbed leads to design and evaluation of pro-\ntocols for heterogeneous wireless networks.\n■\nUCSB MeshNet [49]: The University of California, Santa Barbara, de-\nployed an experimental testbed on their campus. It consists of 25\nnodes equipped with multiple IEEE 802.11a/b/g wireless radios. The\nmain objective of the testbed is to design protocols for the robust\noperation of multi-hop wireless networks. Specifically, the testbed\nis being used to conduct research on scalable routing protocols,\nefficient network management, multimedia streaming, and QoS for\nmulti-hop wireless networks.\n1.5.5\nIndustrial Research in WMNs\nMany companies started research in WMNs on their own and in collab-\noration with academic research institutions. Some of them recently came\nup with mesh networking products for implementing mesh network-based\napplications. In this section some of the industries working toward research\naspects of WMNs and some of the industries providing mesh networking\nproducts are discussed.\n■\nMicrosoft Research [50]: Microsoft researchers at Redmond, Cam-\nbridge, and Silicon Valley are working to create wireless technolo-\ngies that allow neighbors to connect their home networks together\n(community networking). They deployed their own mesh network\ntestbed in their office building and local apartment complex. They\ndeveloped a software module called the Mesh Connectivity Layer\n(MCL) which implements ad hoc routing and link quality measure-\nment. Architecturally, MCL is a loadable Windows driver. It imple-\nments a virtual network adapter, so that the ad hoc network appears\nas an additional (virtual) network link to the rest of the system.\nThe routing protocol used by MCL is LQSR, which improves net-\nwork performance by supporting link-quality metrics for routing.\nThe MCL driver implements an interposition layer between the link\nlayer and the network layer. To higher-layer software, MCL appears\nto be just another Ethernet link, albeit a virtual link. To lower-layer\nsoftware, MCL appears to be just another protocol running over the\nphysical link. This design has several significant advantages. First,\nhigher-layer software runs unmodified over the ad hoc network.\nThe testbed runs both IPv4 and IPv6 over the ad hoc network with-\nout requiring any modifications to the network layer. All network\nlayer functionalities such as ARP, DHCP, and Neighbor Discovery\nwork well. Second, the ad hoc routing runs over heterogeneous link\n" }, { "page_number": 50, "text": "An Introduction to Wireless Mesh Networks\n■\n39\nlayers as well. This implementation supports Ethernet-like physical\nlink layers (e.g., 802.11 and 802.3), but the architecture accommo-\ndates link layers with arbitrary addressing and framing conventions.\nThe virtual MCL network adapter can multiplex several physical net-\nwork adapters, so the ad hoc network can be extended across het-\nerogeneous physical links. Third, the design can support other ad\nhoc routing protocols as well.\n■\nIntel [51]: A wide variety of research and development efforts at Intel\nare geared toward understanding and addressing the technical chal-\nlenges for realizing multi-hop mesh networks. Intel’s Network Archi-\ntecture Lab is aimed at overcoming many of the challenges faced by\nWMNs. They developed low-cost and low-power AP prototypes or\nnodes to enable further research on security, traffic characterization,\ndynamic routing and configuration, and QoS problems. Intel is also\nworking with other industries to develop standards and protocols\nthat support WMNs and enable interoperability between products\nfrom multiple vendors. Intel is working to simplify the entire instal-\nlation process, including network node placement and configuration\nso that end users and businesses can easily realize the full benefits\nof multi-hop mesh networking.\n1.5.6\nMesh Networking Products\n■\nStrix Systems [52]: The mesh networking products from Strix Systems\nare RF-independent supporting existing wireless standards 802.11a/\nb/g and 802.16 (WiMAX), designed to easily add in any future wire-\nless technologies. The Strix Access/One® family of products delivers\nhigh-performance WMN systems by employing modular future-proof\narchitecture supporting multi-radio, multi-channel, and multi-RF\nmesh networking technologies. The Access/One architecture deliv-\ners the industry’s most scalable and flexible wireless networking\nplatform by which the largest citywide and countrywide communica-\ntion services can be built. Unlike competing single and other multi-\nradio products, the Access/One design makes secure full-duplex\ntransmission, instant path switching, and application classification a\nreality. Strix Access/One networks are deployed in many different\nenvironments and used for many different applications around the\nworld, enabling users to access wireless broadband applications at\nany place, anywhere, any time even while moving at 200 miles per\nhour. Strix Access/One is a scalable self-configuring and self-healing\nsystem designed to meet the needs of service providers, government\nagencies, and outdoor mobile enterprises.\n■\nNortel [53]: Nortel’s WMN solution addresses the market require-\nments for networks that are highly scalable and cost-effective,\n" }, { "page_number": 51, "text": "40\n■\nSecurity in Wireless Mesh Networks\noffering end user security, seamless roaming beyond traditional\nWLAN boundaries, and provides easy deployment in areas that do\nnot (or cannot) support a wired backhaul. Nortel’s WMN solution\nis well-suited for providing broadband wireless access in areas that\ntraditional WLAN systems are unable to cover. Nortel provides a\nnumber of products for WMN solutions, which include wireless AP,\nwireless bridge, WLAN security switches, and enterprise network\nmanagement system. These products provide a number of applica-\ntions for the mobile users such as secure mobile networking and\nvoice connectivity featuring flexible seamless mobility across cam-\npus environments, IP telephony and converged multimedia applica-\ntions, and low-cost, high-capacity point-to-point broadband trans-\nmission.\n■\nKiyon Mesh Network [54]: Kiyon also provides mesh networking\nproducts for realizing WMNs. The KAN254B wireless BACNet router\nprovides a WMN solution to industry and converts all standard field\ncontrollers or supervisory controllers using BACnet MSTP, BACnet\nIP, or Ethernet IP to a WMN. It can also be used for security sys-\ntems, video cameras, lighting systems, fire, and Internet applications.\nPeople have applied them in offices and warehouses and even to\nconnect buildings together when running wires was prohibitive.\n■\nFireTide® [55]: FireTide mesh networking provides solutions to ed-\nucation, health care, hospitality, municipal government, and ware-\nhousing. The mesh networking products from FireTide such as\nHotspot indoor and outdoor mesh nodes provide a high-capacity\nwireless mesh backbone for outdoor and indoor networks. These\nproducts are designed for maximum performance, scalability, and\nease of use. They can operate in 2.4- and 5-GHz frequency spectrum.\nThe public safety mesh nodes are ideal for public safety agencies.\nThis operates in 4.940- to 4.990-GHz spectrum, which has been al-\nlocated for public safety agencies in the United States.\n1.6\nSummary\nWMNs have emerged as a promising technology for next-generation net-\nworking. In WMNs, no cabling is required to connect the mesh routers.\nAll mesh routers self-configure wirelessly to form a rich radio mesh back-\nbone network. The wireless connectivity between routers significantly re-\nduces the deployment and maintenance cost when compared with wired\nnetworks. Due to these attractive features of WMNs, they are considered\nfor a wide variety of applications such as community networking, emer-\ngency operations, home networking, and hybrid wireless architectures. In\nthis chapter, the major issues and applications of WMNs were described.\n" }, { "page_number": 52, "text": "An Introduction to Wireless Mesh Networks\n■\n41\nThe design issues and deployment scenarios were also discussed. Provid-\ning high throughput is the major design goal of WMNs, which has been\naddressed in multiple layers. To improve the performance of WMNs, the\nmulti-channel, multi-radio architecture has been suggested. The related pro-\ntocols for this architecture in MAC and routing layer were discussed. Some\nrouting metrics were described to find high-throughput paths by taking\ninto account the channel quality and inter flow and intra flow interference.\nSecurity and standardization are the main concerns for the wide deploy-\nment of WMNs. Some of the security issues and standards such as IEEE\n802.11s and IEEE 802.16 mesh were also discussed. Finally, to provide in-\nsight on real implementations of WMNs, some WMN testbeds and mesh\nnetworking products were also discussed.\nReferences\n[1]\nI. F. Akyildiz, X. Wang, and W. Wang, Wireless mesh networks: A survey,\nComputer Networks Journal, vol. 47, no. 4, pp. 445–487, March 2005.\n[2]\nP. Gupta and P. R. Kumar, The capacity of wireless networks, IEEE Trans-\nactions on Information Theory, vol. 46, no. 2, pp. 388–402, March 2000.\n[3]\nIEEE 802.11 Standard Group Website, http://www.ieee802.org/11/\n[4]\nIEEE Std 802.11-1997, Part 11: Wireless LAN medium access control (MAC)\nand physical layer (PHY) specifications, The Institute of Electrical and Elec-\ntronics Engineers, 1997.\n[5]\nIEEE Std 802.11a-1999, Part 11: Wireless LAN medium access control (MAC)\nand physical layer (PHY) specifications: High-speed physical layer in the 5\nGHz band, The Institute of Electrical and Electronics Engineers, 1999.\n[6]\nIEEE Std 802.11b-1999, Part 11: Wireless LAN medium access control (MAC)\nand physical layer (PHY) specifications: Higher-speed physical layer ex-\ntension in the 2.4 GHz Band, The Institute of Electrical and Electronics\nEngineers, 1999.\n[7]\nP. Kyasanur and N. H. Vaidya, Capacity of Multi-Channel Wireless Net-\nworks: Impact of Number of Channels and Interfaces, Proceedings of ACM\nMOBICOM 2005, pp. 43–57, August 2005.\n[8]\nI. Emre Telatar, Capacity of multi-antenna Gaussian channels, European\nTransactions on Telecommunications, vol. 10, no. 6, pp. 585–595, Novem-\nber/December 1999.\n[9]\nV. Bharghavan, A. Demers, S. Shenker, and L. Zhang, MACAW: A Media\nAccess Protocol for Wireless LANs, Proceedings of ACM SIGCOMM 1994,\npp. 212–225, August 1994.\n[10]\nC. L. Fullmer and J. J. Garcia-Luna-Aceves, Floor Acquisition Multiple Access\nProtocol for Wireless LANs, Proceedings of ACM SIGCOMM 1995, pp. 262–\n273, August 1995.\n[11]\nF. A. Tobagi and L. Kleinrock, Packet switching in radio channels: Part II:\nThe hidden terminal problem in carrier sense multiple access and the busy\ntone solution, IEEE Transactions on Communications, vol. 23, no. 12, pp.\n1417–1433, December 1975.\n" }, { "page_number": 53, "text": "42\n■\nSecurity in Wireless Mesh Networks\n[12]\nF. Talucci and M. Gerla, MACA-BI (MACA by Invitation): A Wireless MAC\nProtocol for High Speed Ad Hoc Networking, Proceedings of IEEE ICUPC\n1997, vol. 2, pp. 913–917, October 1997.\n[13]\nS. Jiang, J. Rao, D. He, and C. C. Ko, A simple distributed PRMA for MANETs,\nIEEE Transactions on Vehicular Technology, vol. 51, no. 2, pp. 293–305,\nMarch 2002.\n[14]\nZ. Tang, and J. J. Garcia-Luna-Aceves, A Protocol for Topology-Dependent\nTransmission Scheduling in Wireless Networks, Proceedings of IEEE WCNC\n1999, vol. 3, no. 1, pp. 1333–1337, September 1999.\n[15]\nZ. Tang and J. J. Garcia-Luna-Aceves, Hop-Reservation Multiple Access\n(HRMA) for Ad Hoc Networks, Proceedings of IEEE INFOCOM 1999, vol. 1,\npp. 194–201, March 1999.\n[16]\nB. S. Manoj and C. Siva Ram Murthy, Real-Time Traffic Support for Ad Hoc\nWireless Networks, Proceedings of IEEE ICON 2002, pp. 335–340, August\n2002.\n[17]\nV. Kanodia, C. Li, A. Sabharwal, B. Sadeghi, and E. Knightly, Ordered Packet\nScheduling in Wireless Ad Hoc Networks: Mechanisms and Performance\nAnalysis, Proceedings of ACM MOBIHOC 2002, pp. 58–70, January 2002.\n[18]\nI. Karthikeyan, B. S. Manoj, and C. Siva Ram Murthy, A distributed laxity-\nbased priority scheduling scheme for time-sensitive traffic in mobile ad hoc\nnetworks, Ad Hoc Networks Journal, vol. 3, no. 1, pp. 27–50, January 2005.\n[19]\nV. Kanodia, C. Li, A. Sabharwal, B. Sadeghi, and E. Knightly, Distributed\npriority scheduling and medium access in ad hoc networks, ACM/Baltzer\nJournal of Wireless Networks, vol. 8, no. 5, pp. 455–466, September 2002.\n[20]\nJ. So and N. H. Vaidya, Multi-Channel MAC for Ad Hoc Networks: Handling\nMulti-Channel Hidden Terminals Using a Single Transceiver, Proceedings\nof ACM MOBIHOC 2004, pp. 222–233, May 2004.\n[21]\nP. Bahl, R. Chandra, and J. Dunagan, SSCH: Slotted Seeded Channel Hop-\nping for Capacity Improvement in IEEE 802.11 Ad Hoc Wireless Networks,\nProceedings of ACM MOBICOM 2004, pp. 216–230, September 2004.\n[22]\nA. Adya, P. Bahl, J. Padhye, A. Wolman, and L. Zhou, A Multi-Radio Uni-\nfication Protocol for IEEE 802.11 Wireless Networks, Proceedings of IEEE\nBROADNETS 2004, pp. 344–354, October 2004.\n[23]\nP. Kyasanur and N. H. Vaidya, Routing and Interface Assignment in Multi-\nChannel Multi-Interface Wireless Networks, Proceedings of IEEE WCNC\n2005, vol. 4, pp. 2051–2056, March 2005.\n[24]\nP. Kyasanur and N. H. Vaidya, Routing and link-layer protocol for multi-\nchannel multi-interface ad hoc wireless networks, ACM Mobile Computing\nand Communications Review, vol. 10, no. 1, pp. 31–43, January 2006.\n[25]\nC. E. Perkins and E. M. Royer, Ad Hoc On-Demand Distance Vector Routing,\nProceedings of IEEE Workshop on Mobile Computing Systems and Applica-\ntions, pp. 90–100, February 1999.\n[26]\nD. B. Johnson, D. A. Maltz, and J. Broch, DSR: The Dynamic Source Rout-\ning Protocol for multi-hop wireless ad hoc networks, Ad Hoc Networking,\nChapter 5, pp. 139–172, Addison-Wesley, 2001.\n[27]\nR. Ogier, F. Templin, and M. Lewis, Topology dissemination based on\nreverse-path forwarding (TBRPF), IETF RFC 3684, February 2004.\n" }, { "page_number": 54, "text": "An Introduction to Wireless Mesh Networks\n■\n43\n[28]\nD. S. J. D. Couto, D. Aguayo, J. Bricket, and R. Morris, A High-Throughput\nPath Metric for Multi-Hop Wireless Routing, Proceedings of ACM MOBICOM\n2003, pp. 134–146, September 2003.\n[29]\nD. Aguayo, J. Bicket, S. Biswas, G. Judd, and R. Morris, Link-Level Mea-\nsurements from an 802.11b Mesh Network, Proceedings of ACM SIGCOMM\n2004, pp. 121–132, August 2004.\n[30]\nR. Draves, J. Padhye, and B. Zill, Comparison of Routing Metrics for Static\nMulti-Hop Wireless Networks, Proceedings of ACM SIGCOMM 2004, pp.\n133–144, August 2004.\n[31]\nR. Draves, J. Padhye, and B. Zill, Routing in Multi-Radio, Multi-Hop Wire-\nless Mesh Networks, Proceedings of ACM MOBICOM 2004, pp. 114–128,\nSeptember 2004.\n[32]\nY. Yang, J. Wang, and R. Kravets, Interference-Aware Load Balancing for\nMulti-Hop Wireless Networks, Technical Report UIUCDCS-R-2005-2526,\nDepartment of Computer Science, University of Illinois at Urbana-\nChampaign, 2005.\n[33]\nG. Hallond and N. Vaidya, Analysis of TCP Performance over Mobile Ad\nHoc Networks, Proceedings of ACM MOBICOM 1999, pp. 219–230, August\n1999.\n[34]\nK. Chandran, S. Raghunathan, S. Venkatesan, and R. Prakash, A feedback\nbased scheme for improving TCP performance in ad hoc wireless net-\nworks, IEEE Personal Communications Magazine, vol. 8, no. 1, pp. 34–39,\nFebruary 2001.\n[35]\nJ. Liu and S. Singh, ATCP: TCP for mobile ad hoc networks, IEEE Journal on\nSelected Areas in Communications, vol. 19, no. 7, pp. 1300–1315, July 2001.\n[36]\nChi-Fu Huang, Hung-Wei Lee, and Yu-Chee Tseng, A two-tier heteroge-\nneous mobile ad hoc network architecture and its load-balance routing\nproblem, Mobile Networks and Applications, vol. 9, no. 4, pp. 379–391,\nAugust 2004.\n[37]\nI. F. Akyildiz, J. McNair, J. S. M. Ho, H. Uzunalioglu, and W.Wang, Mobility\nManagement in Next Generation Wireless Systems, Proceedings of the\nIEEE, vol. 87, no. 8, pp. 1347–1385, August 1999.\n[38]\nI. F. Akyildiz, J. Xie, and S. Mohanty, A survey of mobility manage-\nment in next-generation all-IP-based wireless systems, IEEE Wireless\nCommunications, vol. 11, no. 4, pp. 16–28, August 2004.\n[39]\nIEEE 802.15 Standard Group Website, http://www.ieee802.org/15/\n[40]\nIEEE 802.16 Standard Group Website, http://www.ieee802.org/16/\n[41]\nIEEE Std 802.16a-2003 (amendment to IEEE Std 802.16-2001), Part 16: Air\ninterface for fixed broadband wireless access systems — Amendment 2:\nMedium access control modifications and additional physical layer specifi-\ncations for 2-11 GHz, The Institute of Electrical and Electronics Engineers,\n2003.\n[42]\nIEEE Std 802.16e-2005, Part 16: Air interface for fixed and mobile broad-\nband wireless access system — Amendment 2: Physical and medium\naccess control layers for combined fixed and mobile operation in licensed\nbands and corrigendum 1, The Institute of Electrical and Electronics\nEngineers Inc., 2006.\n" }, { "page_number": 55, "text": "44\n■\nSecurity in Wireless Mesh Networks\n[43]\nIEEE Std 802.16-2004 (Revision of IEEE Std 802.16-2001), Part 16: Air\ninterface for fixed broadband wireless access systems, The Institute of\nElectrical and Electronics Engineers Inc., 2004.\n[44]\nRoofnet, http://pdos.csail.mit.edu/roofnet/doku.php?id=roofnet\n[45]\nJ. Bicket, D. Aguayo, S. Biswas, and R. Morris, Architecture and Evaluation\nof an Unplanned 802.11b Mesh Network, Proceedings of ACM MOBICOM\n2005, pp. 31–42, August 2005.\n[46]\nD. Aguayo, J. Bicket, S. Biswas, D. S. J. De Couto, and R. Morris, MIT\nRoofnet Implementation, http://pdos.csail.mit.edu/roofnet/design/\n[47]\nUCSD Mesh Networks Testbed, http://www.calit2.net/\n[48]\nWireless Mesh Networks, http://www.ece.gatech.edu/research/labs/bwn/\nmesh/\n[49]\nUCSB MeshNet, http://moment.cs.ucsb.edu/meshnet/\n[50]\nSelf-Organizing Neighborhood Wireless Mesh Networks, http://research.\nmicrosoft.com/mesh/\n[51]\nMulti-Hop\nMesh\nNetworks,\nhttp://www.intel.com/technology/comms/\ncn02032.htm\n[52]\nStrix Systems, http://www.strixsystems.com/\n[53]\nWireless Mesh Network Solution, http://www.nortel.com\n[54]\nKiyon, http://www.kiyon.com/\n[55]\nFiretide Instant Mesh Network, http://www.firetide.com/\n" }, { "page_number": 56, "text": "Chapter 2\nMesh Networking in\nWireless PANs, LANs,\nMANs, and WANs\nNeila Krichene and Noureddine Boudriga\nContents\n2.1\nIntroduction ...........................................................47\n2.2\nWireless Mesh Networking Fundamentals ...........................48\n2.2.1\nNetwork Architecture .........................................48\n2.2.2\nCharacteristics .................................................49\n2.2.3\nSupported Applications........................................50\n2.2.4\nRouting Protocols ..............................................51\n2.2.5\nNetwork Management ........................................52\n2.2.6\nQoS Provision .................................................53\n2.2.7\nSecurity Considerations ........................................55\n2.2.8\nScheduling and Multimedia Support .........................55\n2.3\nWireless Mesh PANs ...................................................56\n2.3.1\nBackground and Objectives ..................................56\n2.3.2\nChallenges .....................................................56\n2.3.3\nArchitecture ....................................................57\n2.3.4\nThe IEEE 802.15.5 Standard ..................................57\n2.3.4.1\nMeshing and the Ultra Wide Band ..................58\n2.3.4.2\nOverview of the ZigBee IEEE 802.15.4 Standard ...59\n2.3.4.3\nIEEE 802.15.4 Physical Layer.........................59\n45\n" }, { "page_number": 57, "text": "46\n■\nSecurity in Wireless Mesh Networks\n2.3.4.4\nIEEE 802.15.4 MAC Layer ............................59\n2.3.4.5\nOverview of the IEEE 802.15.5 Standard ...........60\n2.3.4.6\nRouting and QoS Support ...........................62\n2.4\nWireless Mesh LAN ....................................................65\n2.4.1\nIntroduction and Advantages .................................65\n2.4.2\nArchitecture Technologies .....................................67\n2.4.3\nChallenges .....................................................68\n2.4.4\nThe IEEE 802.11s Standard ....................................68\n2.4.4.1\nIEEE 802.11s Device Classes .........................69\n2.4.4.2\nMedium Access Control: The Medium\nAccess Coordination Function .......................70\n2.4.5\nRouting and QoS Support ....................................73\n2.4.5.1\nWMR Protocol Overview ............................73\n2.4.6\nOverview of Available Commercial Systems .................77\n2.5\nWireless Mesh MAN ...................................................78\n2.5.1\nPurpose ........................................................78\n2.5.2\nTargeted Services .............................................78\n2.5.3\nArchitecture ....................................................79\n2.5.4\nStandards ......................................................80\n2.5.4.1\nMAC Layer Overview in WiMAX Mesh Mode ......81\n2.5.4.2\nHand-Over............................................85\n2.5.4.3\nPhysical Layer Overview in WiMAX\nMesh Mode ..........................................86\n2.5.4.4\nQoS Support ..........................................86\n2.5.5\nDeployed Solutions ...........................................90\n2.5.5.1\nTropos® Networks ...................................90\n2.5.5.2\nStrix Systems .........................................93\n2.6\nWireless Mesh WAN ...................................................94\n2.6.1\nIEEE 802.16 Mobility Management ............................95\n2.6.2\nIEEE 802.20 ....................................................96\n2.6.2.1\n802.20 PHY Layer ....................................96\n2.6.2.2\n802.20 MAC Layer ...................................97\n2.7\nAdvanced Issues.......................................................99\n2.7.1\nPhysical Layer .................................................99\n2.7.2\nMAC Layer .....................................................99\n2.7.3\nNetwork Layer .............................................. 100\n2.7.4\nTransport Layer ............................................. 100\n2.7.5\nApplication Layer ........................................... 101\n2.7.6\nNetwork Management ...................................... 102\n2.7.7\nSecurity ...................................................... 102\n2.8\nConclusion .......................................................... 103\nReferences................................................................. 104\n" }, { "page_number": 58, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n47\nWireless mobile mesh networks are made up by several mobile nodes,\nfully wirelessly interconnected, which adopt multi-hop communication for\ndata transmission. This chapter intends to argue why mesh networking\ntechnology represents a new issue to address for wireless networks by pre-\nsenting the mesh networking fundamentals in wireless PANs, LANs, MANs,\nand WANs. For this purpose, we will first study the mesh networking charac-\nteristics while stressing the targeted applications, the network architecture,\nand the particularities of the routing, quality of service (QoS) provision, and\nmanagement protocols. Then, details of the IEEE standardization efforts tar-\ngeting the network coverage ranging from PANs to WANs are presented.\nWe conclude by presenting some of the deployed solutions and discussing\nadvanced design issues aiming at providing scalable, low-cost, and easily\ndeployable Wireless Mobile Mesh Networks.\n2.1\nIntroduction\nThe mobile ad hoc networks (or MANET) have gained researchers’ atten-\ntion for 30 years [1]. MANET nodes share wireless links and can play the\nrole of client and router at the same time without relying on any infras-\ntructure; thus accomplishing large deployment ease and investments cost\ndecrease. Besides, the ephemeral nature of MANETs particularly copes with\ncritical applications such as disaster recovery and battlefield communica-\ntions. Many research works have addressed the multi-hop communication\nissue in wireless networks, but the practical impact was not very impor-\ntant because users rarely operate in ad hoc mode. For instance, the targeted\napplications were limited to specialized missions inducing an unreasonable\ncost, while users searched mostly for cheap information sharing and Inter-\nnet access. Client satisfaction has created a new research topic that aims at\nrevising the MANET concept by considering the MANET network as a flex-\nible and low-cost extension of wired infrastructure networks that integrates\nthem. As a result, the wireless mesh networking paradigm, which inherits\nsome MANET characteristics and targets civilian applications, was born.\nIt is worth noticing that both the wired Internet and the public switched\ntelephone network may be classed as mesh networks [2]; however, future\nwireless mesh networks should rely on a wireless infrastructure to inter-\nconnect mobile devices in a multi-hop fashion. Wireless mesh networks\n(WMNs) support home and enterprise networking applications; they also\nprovide ubiquitous Internet access and enable the implementation of intelli-\ngent transportation systems and public safety applications. Besides, their\ndeployment does not require important investments comparable to the\ndeployment of wired solutions. In fact, wireless mesh routers can rapidly\n" }, { "page_number": 59, "text": "48\n■\nSecurity in Wireless Mesh Networks\nand easily integrate the wireless infrastructure as soon as the coverage\nneeds to be extended. As a result, a growing number of cities have adopted\nthis paradigm to attract visitors and citizens and start a long-lasting devel-\nopment process. Users can temporarily join the mesh network and act as\nclients and routers for other nodes, thus enhancing the network capacity,\nthroughput, and reliability. Currently, one can find off-the-shelf and propri-\netary mesh networks solutions while IEEE standardization efforts are target-\ning network coverage ranging from PANs to WANs. The goal of this chapter\nis to present the mesh networking fundamentals in wireless PANs, LANs,\nMANs, and WANs. To this end, a general overview of the mesh networks\narchitecture and characteristics is given while addressing general concepts\nsuch as the supported applications, the routing and management protocols,\nthe QoS provision, and the security considerations. Then, the detail of the\nIEEE standardization efforts targeting the network coverage ranging from\nPANs to WANs is presented. We particularly address the physical layer and\nthe MAC layer design issues for the mesh communication mode support\nwhile presenting the challenges that are particular to each network (PAN,\nLAN, MAN or WAN). An overview of the available commercial systems and\ndeployed solutions is also given. We conclude by discussing some of the re-\nsearch issues aiming at designing scalable, low-cost, and easily deployable\nwireless mobile mesh networks.\n2.2\nWireless Mesh Networking Fundamentals\n2.2.1\nNetwork Architecture\nA wireless mesh network is a hierarchical network formed by fully wire-\nlessly interconnected nodes, as illustrated in Figure 2.1. A fully meshed\nnetwork is a network where every node directly connects to every other\nnode; a partial mesh network is a network where each node is connected\nto a set of other nodes [47]. We distinguish routers nodes that act as layer 3\ngateways and support meshing functions. Such nodes are usually equipped\nwith multiple network interfaces for different access technologies; they can\nguarantee wider coverage with less power consumption thanks to the sup-\nport of multi-hop communications. The network resulting from the mesh\nrouters interconnection is called a wireless backbone; it guarantees the con-\nnectivity between nomadic users and wired gateways. The wireless mesh\nnetwork includes also Access Points (APs), which can be viewed as special\nmesh routers provided with a high-bandwidth wired connection to the\nInternet. The wireless network formed by the interconnection of the AP\nand the mesh routers is called a backhaul. The latter enables the access to\nexternal networks while providing high-bandwidth and seamless multi-hop\ncommunication at a low cost.\n" }, { "page_number": 60, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n49\nWired Internet\nbackbone\nAccess \npoints \nWireless \nrouters \nNomadic \nusers \nWired/Wireless connection \nWireless connection \nFigure 2.1\nThe wireless mesh network architecture.\nFinally, mesh clients are generally equipped with a radio interface sup-\nporting mesh networking functions; that is why they can act as routers\nfor other mesh nodes. However, they do not provide the bridge/gateway\nfunctionalities needed for Internet access and interoperability with other\nnetworking technologies. Mesh clients can be laptops, pocket PCs, PDAs,\nIP phones, etc.\n2.2.2\nCharacteristics\nMesh networks are gaining a growing interest thanks to their special char-\nacteristics that enable the deployment of new applications at lower cost.\nThe most important characteristics are as follows:\n■\nMulti-Hop Communication: The multi-hop communication scheme\nguarantees larger coverage zones and an enhancement of the net-\nwork capacity. In fact, line-of-sight constraint no longer matters\nbecause the intermediate nodes relay the information to their neigh-\nbors on short wireless links using a reduced power transmission.\nAs a result, the interferences are decreased and the throughput is\n" }, { "page_number": 61, "text": "50\n■\nSecurity in Wireless Mesh Networks\naugmented [3]. Besides, the multi-hop connectivity allows several\ndevices to access the network at once by relying on other mesh\nnodes without affecting the overall network performance. Finally,\nmesh networks gain more capacity as the number of internal nodes\nincreases and the data traffic can reach larger areas by crossing mul-\ntiple hops until the final destination.\n■\nWide Coverage and Cost Reduction: The wireless infrastructure sup-\nported by the mesh networks eliminates the deployment costs of\na new wired backhaul through cities and rural areas. Moreover,\nthe flexible infrastructure can easily be enforced by adding new\nwireless mesh routers anywhere, anytime the coverage needs to be\nenhanced. Only some APs need to be connected with the wired\ninfrastructure to allow Internet access.\n■\nSelf-Configuration and Self-Management: New mesh nodes that\nenter the network are transparently supported because meshing\nfunctions such as neighbors discovery and automatic topology learn-\ning are implemented. Wireless routers rapidly detect the presence of\nnew paths, thus enhancing the overall performance and coverage.\n■\nNetwork Access and Interoperability: Backhaul devices are equipped\nwith multiple network interfaces that support both Internet and peer-\nto-peer communications while guaranteeing access to existing wire-\nless networks technologies such as traditional IEEE 802.11, WiMAX,\nZigBeeTM, and cellular networks.\n■\nMobility and Power Consumption: The mobility and power consum-\nption vary with the nature of the mesh node. For example, mesh\nrouters and APs have minimal mobility and reduced power con-\nstraints. However, mesh clients are mostly small mobile devices\nwith reduced battery autonomy. Therefore, MAC and routing proto-\ncols supported by the backbone/backhaul do not need to be power\nefficient, but they cannot be implemented on simple mesh clients.\n■\nReliability: Mesh networks rely on multi-hop communication and\ncan use every internal node to route traffic to the destination. There-\nfore, multiple paths exist between two communicating endpoints\nand temporary path failures can be easily tolerated. Besides, mesh\nclients that need to communicate with external destinations (e.g.,\nInternet) can choose between multiple egress points toward the\nwired network, thus tolerating router failures and reducing potential\ncongestions.\n2.2.3\nSupported Applications\nThe mesh networks support a large number of applications dedicated to\npersonal, local, metropolitan, and wide areas networks.\n" }, { "page_number": 62, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n51\n■\nHome Networking: Mesh networks can be deployed at home\nbecause they support bandwidth-greedy applications such as multi-\nmedia traffic transmission [5]. Mesh nodes can be desktop PCs, lap-\ntops, high-definition TV, and DVD players. Wireless APs or mesh\nrouters can easily be added to cover dead zones without requiring\nwiring or complex configurations.\n■\nEnterprise Networking: Traditional wireless LANs have been widely\nused in enterprises, but they have not succeeded in effectively reduc-\ning the deployment cost because the presence of a wired infrastruc-\nture is a must. Adopting mesh networks in enterprises enables the\nshare of resources and an overall performance enhancement thanks\nto the multi-hop communication and the wireless infrastructure de-\nployment. In fact, bottleneck congestion resulting from the one-hop\naccess to the traditional APs is eliminated. Besides, the infrastructure\ncan easily scale according to the network’s needs without requiring\ncomplex configurations and wiring.\n■\nPublic Applications: Mesh networks support public applications at\nthe metropolitan and wide area scale mainly because the line-of-\nsight constraint can be overcome. Wireless Internet access on the\nroad, public safety, and implementation of intelligent transportation\nsystems are highly appreciated by cities’ inhabitants and visitors, and\nhave already been deployed in many countries such as the United\nStates, Taiwan, and Bangladesh.\nThe supported applications will be further detailed later in the chapter.\n2.2.4\nRouting Protocols\nWireless mesh networks are characterized by multi-hop communications\nand rely on a wireless backhauling system to access other external net-\nworks such as the Internet. Consequently, they need to address special\nconstraints such as enhanced scalability, varying power constraints, and\ncross-layer design. These specificities require special routing capabilities\nthat may be partially inherited from the ad hoc context, but that surely differ\nfrom those implemented in the wired and cellular networks. We believe that\nthe specification of a wireless mesh routing protocol should provide new\nperformance metrics that take into consideration the quality of the inter-\nmediate links while trying to minimize the path length. Meanwhile, the\nmesh routers and the mesh clients presenting different mobility and power\nconstraints should implement an efficient hybrid routing protocol able to\naddress those specificities. For instance, the Link Quality Source Routing\n(LQSR) based on the DSR protocol [49] selects the routes with respect to\nthe expected transmission count (or ETX, [52]), the per-hop round-trip tune\n(RTT), and the per-hop packet pair. Results showed that adopting the ETX\n" }, { "page_number": 63, "text": "52\n■\nSecurity in Wireless Mesh Networks\nfor stationary nodes guarantees a good performance although adopting\nthe minimum hop count as route selection criteria for mobile nodes gives\nbetter results. New performance metrics that achieve good performances\nin the mesh context present a research issue that needs to be investigated.\nIn addition, fault-tolerance mechanisms that guarantee the rapid selec-\ntion of a new path in case of link failure should be defined. Besides, the\nroute selection should be based on the congestion status of the network to\nefficiently use the available resources. In fact, the mesh network presents\nmultiple routes between communicating nodes so that alternative paths\nwhich offer the required QoS may be selected in case of mobility or link\nquality decrease. However, it is worth noticing that the route-establishment\ncomplexity increases as the network size grows. Meanwhile, the rout-\ning protocol should address the ephemeral nature of mesh nodes while\nguaranteeing the end-to-end QoS requirements, especially in the case of\nmetropolitan and wide area mesh networks. When considering the ad hoc\ncontext, hierarchical routing protocols as presented in [53–55] adopt a self-\norganization scheme that groups the network nodes into clusters with a\ncertain size. Each cluster is then managed by one or more clusterheads and\nnodes belonging to different clusters may communicate using other nodes\nas gateways. The routing mechanisms implemented inside a cluster may\nbe proactive while intra-cluster routing may be on-demand. Such protocols\nachieve good performances especially when the node’s density is high;\nhowever, they cannot be applied to the mesh context without adding some\nmodifications. For instance, a mesh node selected as a clusterhead may\nnot present sufficient power and processing capabilities, thus becoming\na bottleneck. Geographic routing which is topology-based resists mobility\nbetter, but requires important processing resources. In addition, delivery is\nnot always guaranteed even if a path exists between the communicating\nnodes. Open research issues need to be addressed if this routing principle\nis applied to the mesh networking context.\n2.2.5\nNetwork Management\nMesh networks management needs to address nodes’ specificities in terms\nof mobility, location, and power to provide an up-to-date vision of the net-\nwork status. The resulting accurate management data will serve especially\nfor enhancing the overall performances and making the wise decisions to\novercome the encountered problems.\n■\nMobility\nManagement:\nMobility\nmanagement\naddresses\nthe\nlocation management and the hand-over. Location management ad-\ndresses the location registration and the call delivery; it guarantees\nthat active nodes remain always reachable despite their mobility. The\nhand-over process, also known as hand-off, consists in transferring\n" }, { "page_number": 64, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n53\na communication; therefore, it requires a new connection genera-\ntion and implements the control of the data flow. Advanced mobility\nmanagement mechanisms have been proposed for cellular and IP\nnetworks; however, the adopted schemes are centralized because\nthey rely on the base stations. As mesh networks present an ad\nhoc architecture, distributed or hierarchical location and hand-over\nmanagement functions should be adopted while taking into con-\nsideration the nodes’ nature (routers or clients) and their different\nmobility schemes. In fact, backbone nodes present reduced mobility\nwhile mesh clients frequently roam across different mesh routers.\nProposing a multi-layer mobility management framework that\naddresses mesh specificities is a hot research topic that needs to be\ninvestigated. More specifically, location management functions may\nbe used at MAC and routing layers to provide better performances\nand permit the development of new location-based applications for\nthe mesh scenarios.\n■\nPower Management: Mesh networks are made up of mesh routers\nand mesh clients. While the routers present reduced mobility and\npower constraints, the clients are tiny pieces of equipment, such\nas IP phones and sensors, which are battery-dependent. Besides, it\nis always preferable to reduce the transmission power to save the\nresources and reduce the interferences while increasing the spec-\ntrum spatial-reuse efficiency. Consequently, power-efficient proto-\ncols need to be developed while paying particular attention to some\nconstraints as the hidden nodes scenario to avoid the performance\ndegradations at the MAC level.\n■\nNetwork Monitoring:\nMesh routers need to calculate their own\nstatistics to report them for monitoring servers. Servers should then\nanalyze the data and process anomaly detection. They can then\ntrigger alarms or reactively respond, depending on the scenario. Few\nnetworking management protocols have been proposed for the ad\nhoc context [56]; however, they do not address the scalability issue\nof the mesh networks. Besides, new data processing algorithms that\naddress the mesh network’s specificity need to be developed.\n2.2.6\nQoS Provision\nA service in a communication network is defined by the International\nTelecommunication Union (ITU) as a service provided by the service plane\nto an end user (e.g., a host [end system] or a network element) and which\nutilizes the IP transfer capabilities and associated control and management\nfunctions for delivery of the user information specified by the service level\nagreement (SLA) [69]. In the telecommunications area, the quality of service\n" }, { "page_number": 65, "text": "54\n■\nSecurity in Wireless Mesh Networks\nis intrinsic, perceived, or assessed. Intrinsic QoS is a technical measure con-\nsidered by engineers and network service providers; it is always objectively\ncompared to the expected performance not affected by customers’ percep-\ntions. Perceived QoS reflects the end user’s view about a service while\nassessed QoS is a factor that the customer decides whether or not to con-\ntinue using the service [69]. It is clear that the most challenging issue in\nproviding QoS is to specify the requirements and then quantify them based\non a set of measurable QoS parameters such as the delay, the jitter, and the\nbandwidth.\nToday, most Internet protocols provide best-effort IP forwarding while\nQoS support is required to satisfy multimedia applications needs. To add-\nress this issue, two major QoS models have been proposed: the Integrated\nService (IntServ) [73] and the Differentiated Service (DiffServ) [74]. IntServ\nis a QoS model which adopts virtual circuit connection mechanisms and\noffers per-flow end-to-end reservations. The Resource ReSerVation Protocol\n(RSVP) is used as a signaling protocol to set up and maintain virtual con-\nnections and reserve resources along a route. IntServ provides hard QoS\nguarantees; however, the adopted per-flow granularity leads to a scalability\nproblem because the amount of state information increases with the num-\nber of flows and nodes. DiffServ was designed to overcome the difficulty of\nimplementing and deploying IntServ and RSVP. In fact, the DiffServ scalable\nsolution provides QoS on the wired Internet by defining a set of QoS classes\nand then classifying packets into them according to an SLA negotiated with\nthe Internet Service Provider (ISP). Edge routers perform the complicated\nflows classification while the core routers do not keep per-flow informa-\ntion, but aggregate different packets that were assigned to different classes\non a per-hop behavior (PHB). DiffServ aims to provide service differentia-\ntion among traffic aggregates over a long timescale, but it does not fit to a\nfast topology-changing context.\nQoS routing algorithms deployed in the mesh networks adopt either an\nIntServ or a DiffServ approach according to the network size (coverage area\nand nodes numbers) and the mobility scheme. For instance, MeshDynamics\nproposes a technique for wireless mesh PANs called heartbeats [7], which\nrelies on the information provided by each intermediate node to establish\npaths satisfying the QoS requirements from source to destination. Besides,\n[21] proposes a QoS routing protocol called WMR (Wireless Mesh Routing)\n[21] for a wireless mesh LAN infrastructure. WMR supports multimedia\napplications by guaranteeing minimum bandwidth and maximum end-to-\nend delay for all intra-BSS and inter-BSS communications; it also guarantees\na per-flow granularity and processes a full, on-demand hop-by-hop rout-\ning with no route caching [21]. To fulfill the broadband wireless access\nQoS requirements in MAN networks and address the scalability issues, the\nIEEE 802.16 standard defines four classes of service while [68] presents a\nWireless DiffServ architecture for the wireless mesh backbone.\n" }, { "page_number": 66, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n55\n2.2.7\nSecurity Considerations\nMesh networks need to provide advanced security mechanisms to encour-\nage client subscribing to reliable services. More specifically, the mesh traffic\ntravels through multiple intermediate nodes on the particularly vulnerable\nwireless channels, thus increasing the hacking probability. Currently, mesh\nnetworks provide the same security services deployed in the WLANs and\nencrypt the backhaul communications which represent the important part\nof the whole traffic [4]. However, they have some characteristics that render\nthem particularly vulnerable [6]. In fact, the adopted multi-hop communi-\ncation which relies on the cooperation of the network nodes suffers from\nselfish behaviors. For instance, some selfish nodes may obtain free ser-\nvices while refusing to participate in routing and affecting the system avail-\nability. Besides, the lack of authentication provides attacking nodes with\nfree-of-charge services. Consequently, hackers may cause denial of service\nby sending arbitrary traffic or advertise high rates, thus affecting network\nperformance. Moreover, the routing service which adapts to the topology\nchanges and the environment conditions can be attacked in several ways.\nIn fact, malicious nodes can mislead targeted actors by pretending higher\nor reduced utility values to create an inaccurate representation of the net-\nwork status, thus leading to serious denial of service attacks. To address\nthis issue, each node should locally verify the consistency of the collected\ninformation and base its routing decision on the deduced conclusion.\n2.2.8\nScheduling and Multimedia Support\nMesh networks adopt broadcast scheduling to coordinate transmissions\nbetween the communicating nodes. We mainly distinguish two types of\nscheduling which vary according to the scheduling-messages contention\nresolution procedure [30]. For instance, in the distributed scheduling\nadopted by the IEEE 802.16 standard, the nodes share their scheduling\ndata within the two-hop range and cooperate to avoid contention while\nresources are granted, thanks to a connection establishment procedure.\nHowever, mesh BS collects resource requests from the nodes within a cer-\ntain range and then allocates the resources in a centralized manner [38].\nSuch resource reservation procedures are implemented in the MAC layer to\nestablish high-speed broadband mesh connections needed by multimedia\napplications. In fact, scheduling supplies guaranteed bandwidth and delay\nbased on the flow priority requirements in both metropolitan and wide area\nnetworks [72]. In PAN context, beacons are used to allow isochronous trans-\nmission by reserving Channel Time Allocation (CTA) slots. We may state\nthat the QoS provision mechanisms proposed for mesh networks differ\nfrom one network to another. In the following sections, we further detail\n" }, { "page_number": 67, "text": "56\n■\nSecurity in Wireless Mesh Networks\nthe implementations of MAC and routing aware QoS that intend to support\nmultimedia applications.\n2.3\nWireless Mesh PANs\n2.3.1\nBackground and Objectives\nWireless mesh PANs aim to provide short-range communications between\nsmall groups of fixed and mobile computing devices such as PCs, PDAs,\nperipherals, cell phones, pagers, and consumer electronics. As the network\nnodes have power constraints, the multi-hop communication is adopted to\nincrease the coverage area while reducing transmission power and increas-\ning the throughput. Besides, the nodes do not rely on an infrastructure as\nin wireless LANs; they have to play the role of clients and routers at the\nsame time. Therefore, the network reliability and stability need to be guar-\nanteed despite routers’ mobility. In addition, wireless mesh PANs intend\nto provide multimedia applications that require the design of appropriate\nQoS routing protocols [7]. More specifically, multimedia home networking\nwith high-speed streaming media and streaming content download, envi-\nronmental monitoring, automatic meter reading, and plenty of commercial-\nand industrial-type applications monitoring need to be supported [9].\n2.3.2\nChallenges\nThe reliability of the QoS routing service is a major concern for wireless\nmesh PANs. In fact, in the ad hoc networks context, each node maintains\na connectivity graph defining a path for every other node in the network.\nHowever, the node’s mobility leads to a constant change in the routing\ntables and result in an important overhead as the number of the network\nmembers increases. To address these issues, mesh routings protocols se-\nlect the next relay based on the local information stating which node has\nthe strongest signal and is closest to the sender. Unfortunately, this local\napproach is efficient only in the case of small networks; besides, it is not\nable to guarantee QoS for mission-critical applications. A global approach\nbased on the exchange of compact control messages for the routing tables\nupdates needs to be found. On the other hand, the routing service needs\nto proactively adapt to the power constraints of the nodes to avoid paths\nbreakage and QoS violations. The third wireless mesh PANs challenge is\nrelated to beacon alignment issues. In fact, traditional PANs use beacons\nto provide isochronous transmissions. A beacon is formed by CTA and\nContention Access Period (CAP) time slots, as depicted in Figure 2.2.\nCTA time slots are reserved slots for regular transmissions of traffic with\nhard QoS constraints such as video streaming over a multi-hop network.\n" }, { "page_number": 68, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n57\nNode 7\nNode 1\nB\nCAP\nB\nCAP\n3\n4\nCTA\n2\n5\n4\n1\n2\n8\n6\n9\nFigure 2.2\nTwo beacons experiencing interferences.\nThe Pico Net Controllers (PNCs) send the beacon synchronization pulses\nto coordinate the transmissions between the managed nodes. However,\na node may not receive this pulse due to radio interference from other\ndevices in other pico nets. Consequently, the PNCs should coordinate their\ntransmissions with their managed nodes despite the fact that interference\nmay occur at anytime (during the beaconing period [B], the CAP, or the\nCTA period).\n2.3.3\nArchitecture\nA mesh PAN can either be organized in a full mesh topology or a partial\nmesh topology. When each node is directly connected to all others, we\nobtain a fully meshed network [9]. In a partial mesh topology, only some\nnodes are directly connected to all others; the remaining ones are con-\nnected only to nodes with which they frequently communicate. A mesh\nPAN topology is made up of a PAN Coordinator (PAN-C) that is partially or\nfully connected with other Full Function Devices (FDDs). Each FDD is then\ninterconnected with a set of Reduced Function Devices (RFDs). FDDs sup-\nport enhanced functionalities such as routing and link coordination; RFDs\nare simple send/receive devices. This mesh topology allows better network\ncoverage extension and provides enhanced reliability via route redundancy\nbecause nodes may act as routers and relay data in case of link breakage.\nIn fact, data which has not reached its destination is forwarded to one or\nmore neighbors by nodes that act as repeaters. Each node keeps a routing\ntable that indicates which neighbor to contact when a packet with a par-\nticular address is forwarded. Moreover, an easier network configuration is\nfulfilled and the battery lives are extended due to short links usage.\n2.3.4\nThe IEEE 802.15.5 Standard\nThe IEEE 802.15.5 Working Group was created in May 2004 to define a\ncomplete framework that provides a reliable and scalable wireless connec-\ntivity for mesh nodes based on the specification of the low-rate wireless\n" }, { "page_number": 69, "text": "58\n■\nSecurity in Wireless Mesh Networks\nPANs specified in IEEE 802.15.4 standard and the high-rate wireless PANs\nspecified in IEEE 802.1.5.3 [11,13].\n2.3.4.1\nMeshing and the Ultra Wide Band\nThe Ultra Wide Band (UWB) is a high-speed physical technique that partic-\nularly fits short-range communications. In fact, UWB enhances the meshing\ncapabilities by having low power and cost constraints while guaranteeing\nprecise location information and important throughput. This radio technol-\nogy transmits signals with extremely wide spectrum (e.g., the bandwidth of\nthe transmission can be several GHz wide [18]) at a very low transmission\npower so that the resulting Power Spectrum Density (PSD) is very low, thus\nallowing a massive frequency reuse [10]. For example, 1 W of total power\nspread across 1 GHz of frequency spectrum puts only 1n W of power into\neach hertz band of frequency. The resulting reduction of the consumed\npower allows tiny devices to save their battery life while resisting fading\nand interference. However, UWB applies only to short-range communica-\ntions because the bandwidth decreases rapidly as distance increases [3,10].\nConsequently, if the same throughput offered by the UWB needs to be\nprovided for wireless mesh LANs or MANs, new physical layer transmis-\nsion techniques need to be developed. UWB allows the coexistence of tens\nand even hundreds of simultaneous non-interfering channels within radio\ndistance of each other. Using a mesh topology enables us to trade some\nchannels to increase the overall performance, as illustrated by Figure 2.3.\nIn fact, nodes A and B are direct neighbors distant by 10 m and having\n100 Mbps as available bandwidth. Besides, node C is a common neighbor\ndistant by 5 m from A and B. This shorter distance implies 250 Mbps of\navailable bandwidth between both A and C and B and C. If A wishes to\ncommunicate with B, it will be wise to choose the path A -> C -> B with\nan available bandwidth of 250 Mbps, which is two times faster than the\ndirect one. Meshing also increases the coverage because nodes which are\nnot in direct range can communicate by using other network members as\nrelays. Using large UWB increases the available bandwidth as the number\nof nodes increases. To conclude, the combination of the UWB technology\nC\nB\nA\n5 m/250 Mbps\n10 m/100 Mbps\n5 m/250 Mbps\nFigure 2.3\nMeshing increases the throughput.\n" }, { "page_number": 70, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n59\nand a mesh topology guarantees a very easy and cheap deployment of\ncommunication networks for homes and offices.\n2.3.4.2\nOverview of the ZigBee IEEE 802.15.4 Standard\nThe ZigBee IEEE 802.15.4 standard specifies the PHY and MAC layers imple-\nmentation which intend to support low-rate wireless communications in a\nPAN, which can be either a star, a mesh, or a cluster tree [70]. ZigBee also\naddresses the third layer functionalities and combines tree routing with\non-demand non-tree routing while eliminating single point of failure.\nRoutes forming a tree branch are optimally traced based on the hop\ncount, link quality, and power. Meanwhile, optimal on-demand paths are\northogonal and connect different tree branches. As a consequence, the\ntree routes and the on-demand ones interconnect all the nodes within the\nnetwork and result in a mesh.\nBesides, the network defines three logical devices depending on their\nfunctionalities. In fact, we distinguish the ZigBee coordinator, which is an\nFDD; the ZigBee router that can act as a coordinator within its operating\narea, and the ZigBee end device, which can be either an FDD or an RFD.\nThe mesh topology defined by ZigBee is also known as the peer-to-peer\ntopology. It defines one PAN coordinator, allows any device to communi-\ncate with any other neighboring device, and enables multi-hop transmis-\nsions [70], thus forming an ad hoc self-healing and self-forming network.\n2.3.4.3\nIEEE 802.15.4 Physical Layer\nThe physical layer defines two services: the physical data and the phys-\nical management service. It manages the activation and deactivation of\nthe radio transceiver, the energy detection (ED), the link quality indica-\ntion (LQI), the channel selection, the clear channel assessment (CCA), and\nthe transmitting and reception of packets across the physical medium [70].\nThe adopted modulation technique is the direct sequence spread spectrum\n(DSSS), which offers data rates of 250 kbps at 2.4 GHz, 40 kbps at 915 MHz\nand 20 kbps at 868 MHz. The low frequencies offer an extended range\nwhile the high frequency provides a high throughput. Besides, a single\nchannel is defined between 868 and 868.6 MHz, ten channels are defined\nbetween 902.0 and 928.0 MHz, and 16 channels lie between 2.4 and 2.4835\nGHz, thus enabling channel reallocation within the spectrum. Receiver sen-\nsitivities are −85 dBm for 2.4 GHz and −92 dBm for 868/915 MHz while\nthe maximum transmit confirms with local regulations.\n2.3.4.4\nIEEE 802.15.4 MAC Layer\nThe ZigBee MAC layer provides two services: the MAC data service and the\nMAC management service interfacing to the MAC sub-layer management\n" }, { "page_number": 71, "text": "60\n■\nSecurity in Wireless Mesh Networks\nentity service access point (MLMESAP). The coordinator devises the super-\nframe into 16 equally sized slots and bounds it by network beacons. In fact,\nthe beacon frame is sent in the first slot of each superframe to synchro-\nnize the attached devices, identify the PAN, and describe the superframe\nstructure [70]. Besides, the superframe may have an inactive portion during\nwhich the coordinator enters in a low-power mode and an active portion\nconsisting of the CAP and the contention free period (CFP). Devices that\nwish to communicate during the CAP period need to compete to gain access\nusing a slotted (CSMA/CA) approach. On the other hand, the CFP presents\nguaranteed time slots, which may occupy more than one slot period [70].\nThe beacon is transmitted at the start of slot 0 without the use of CSMA while\nall frames except acknowledgment frames or any data frames that immedi-\nately follow the acknowledgment of a data request command transmitted\nin the CAP shall use slotted CSMA-CA to access the channel. A transmission\nin the CAP shall be complete one IFS period before the end of the CAP,\nwhere an IFS (Inter Frame Space) period is the amount of time necessary to\nprocess the received packet by the physical layer. If the transmission is im-\npossible, it will be deferred until the CAP of the following superframe. The\nCFP starts on a slot boundary immediately following the CAP and extends\nto the end of the active portion of the superframe. Its length is determined\nby the length of the combined guaranteed time slots [70].\n2.3.4.5\nOverview of the IEEE 802.15.5 Standard\nA wireless mesh PAN should guarantee isochronous and asynchronous data\ntransmissions and provide high throughput and low latency while sup-\nporting a high spatial frequency reuse and a decentralized monitoring. To\naddress these issues, [71] proposes the adoption of a superframe with a slot-\nted structure at the MAC layer, as depicted in Figure 2.4. This superframe\nMesh-traffic\nMesh-traffic\nt\nBP\nBP\nSuperframe\nSuperframe\nt\nMedium Access Slot (MAS)\nFigure 2.4\nA 802.15 MAC superframe structure.\n" }, { "page_number": 72, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n61\nBeacon slot number\n6\n5\n4\n3\n2\n1\n0\n7\n8\nDEVID 79\nDEVID 5\nDEVID 14\nDEVID 2\nDEVID 9\nDEVID 38\nBPST\n9\nt\n“m Signal slot count” beacon slots\nreserved for joining devices and\nbeacon collision resolution\nFigure 2.5\nA new device joining the announcement period.\nis made up of multiple Medium Access Slots (MASs) and divided into a\nbeacon period and a mesh traffic period, as shown in Figure 2.4.\nThe beacon period is used to exchange network and topology manage-\nment information while data is transmitted during the mesh traffic period.\nIn fact, each device should transmit a beacon which provides the device\nID and the neighborhood and synchronization information along with the\nneighbors and the medium access information. The beacon size may vary\nand the number of transmitted beacon slots during MAS is determined using\nthe Adaptive Beacon protocol. Several empty beacon slots may be used by\nthe joining devices. In fact, a new device joining the beacon period should\nindicate its presence within the announcement period, as in the case of the\ndevice 38 in Figure 2.5 [71].\nThereafter, the joining node selects one of the available beacon slots,\nas depicted in Figure 2.6. It is worth noticing that the neighbors provide\ninformation about the empty slots and the beacon period duration.\nDuring the beacon period, devices continually listen to the stated infor-\nmation to store the power indication for each beacon and then combine\nthe power and beacon device ID, thus deducing the neighborhood and\ninterference graph.\nBeacon slot available for\njoining devices\nHighest-numbered\noccupied beacon slot\n0\n1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12 13\n14\nt\nDEV 38 chooses\nas one available\nbeacon slot\nDEVID 38\nDEVID 14\nDEVID 5\nDEVID 79\nDEVID 2\nDEVID 9\nBPST\nFigure 2.6\nFinal beacon occupancy.\n" }, { "page_number": 73, "text": "62\n■\nSecurity in Wireless Mesh Networks\nData transmission is scheduled during the data transmission period in\na distributed fashion. Data may be VoIP flows and multimedia streaming\ntransmitted with QoS guarantees. The distributed QoS support is guaran-\nteed by the Distributed Reservation Protocol (DRP), which acts as follows:\ncommunicating devices announce the desired transmissions, the receiver\nand transmitter may negotiate using the beacons, which carry information\non the other reservations. In fact, the transmitter announces its desired\ntransmission with its beacon and the receiver may accept or refuse to com-\nmunicate. High-priority traffic may replace low-priority traffic and data is\ntransmitted in a unidirectional fashion while the interference awareness\nallows parallel transmissions. Small frames are aggregated into larger Proto-\ncol Data Units (PDUs) and may be transmitted to multiple receivers.\n2.3.4.6\nRouting and QoS Support\nThis section presents two different proposals related to routing in wireless\nmesh PANs that have been submitted to the IEEE 802.15.5 Working Group.\n■\nMeshDynamics Proposal: MeshDynamics has submitted a proposal\nfor the IEEE 802.15.5 Work Group that addresses the QoS routing\nissue in wireless mesh PANs. In fact, wireless mesh PANs are char-\nacterized by the mobility of nodes which play the role of routers,\nthus affecting the routing performance and the QoS provision. To\nadapt to the changing topology and the environment conditions, a\ndistributed control layer (Figure 2.7) has been proposed.\nBased on the application requirements in terms of latency and\nthroughput and the nodes’ status and setting in terms of mobility\nMAC level routing\nState of the network\nControl sampling\nMesh routing\nAdaptive latency/Throughput control\nDistributed control layer\nApplication software\nMesh control layer\nMAC-mesh inerface\nMAC\nPHY\n1. Application requirements\n2. Device status & settings\nFigure 2.7\nProposal of a distributed control layer.\n" }, { "page_number": 74, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n63\nand power constraints, a QoS mesh routing is performed. For this\npurpose, the distributed control layer coordinates the mesh routing\nand adapts it to the power status of the nodes. For example, a relay\nnode which needs to enter the sleep mode has just to change its\nmode to a low power setting and send a sleep mode message so\nthat other entities will communicate with it only if it is their final\ndestination. Another path that provides the same QoS but does not\ninclude the sleeping node is then proactively elected.\nTo guarantee a QoS routing service, MeshDynamics proposes a\ntechnique called Heartbeats [7]. For instance, each node within the\nnetwork should send a heartbeat including toll-cost and hop-cost\ninformation, beacon alignment data, link state, and distance vector\ninformation. Each node that has to route a packet should enlist the\nintermediate entities that need to cooperate to guarantee delivery\nwhile providing the required QoS. As intermediate nodes need some-\ntimes to reduce the traffic load when they need to provide better\nservice for the traffic they are generating, they will raise their toll\ncost, which is the cost of using them as relay. Consequently, nodes\nwith higher priorities will pay a higher hop cost for a shorter path\n(lower delay path) with increasing toll cost. Meanwhile, traffic with\nsofter QoS constraints will be routed on longer routes and may\nexperience congestion at popular nodes.\nIn addition, MeshDynamics proposes a software layer on the\nMAC layer that addresses isochronous transmission in Simultaneous\nOperating Piconets by managing the beacon alignment issues with-\nout modifying the MAC IEEE specifications. The principle consists\nof applying a theory to determine if there are common reachable\nnodes that may experience interference. That is, two PNCs that share\na common reachable list of neighbors are not allowed to transmit\nbeacons simultaneously; they should stagger their transmission. A\nPNC that cannot hear any of the other PNCs should hear neighbor-\ning intermediate devices that act as repeaters on behalf of their PNCs\nby sending the heartbeats periodically or as a request response (e.g.,\na node that hears a request asking for location and neighbors’ iden-\ntities sends the last beacon transmitted by its PNC). A more detailed\ndescription of the protocol can be found in [7].\n■\nSamsung Proposal: This sub-section intends to present the Samsung\nproposal for the 802.15.5 wireless mesh PAN targeting the low-rate\nmesh architecture based on the Meshed-Tree approach and address-\ning Meshed Tree routing, multicasting, and key pre-distribution. The\nproposal defines the Adaptive Robust Tree (ART) paradigm, which is\nbased on an adaptive assignment of logical addresses reflecting the\nnetwork topology during the tree definition. The ART defines three\nphases: the initialization (or configuration) phase, the operation\n" }, { "page_number": 75, "text": "64\n■\nSecurity in Wireless Mesh Networks\nphase, and the recovery phase. The initialization phase is triggered\nwhen new nodes join the network and reorganize themselves to\nform the ART. The tree formation requires the execution of two\nsub-phases: the association and the address assigning. Then, each\nnode keeps track of the ART branches in the ART table (ARTT).\nThose branches are assigned one or more blocks of consecutive\nlogical addresses. Communication between nodes starts during the\noperation phase. However, new nodes may integrate the network\nand lead to changes in the topology during this phase; hence many\nreconfigurations may take place to provide an up-to-date status. The\nrecovery phase is triggered when nodes leave the network and cause\nlink breakage. In this case, only the affected tree part is recovered\nwithout changing any assigned address; the other nodes still in the\noperation phase may continue their communications.\nThe ART formation begins by the association stage during which\nnew nodes gradually join the network beginning from the tree root.\nAfter the bottom is reached, a reverse procedure is used to calculate\nthe number of nodes along each branch. After the number of enti-\nties is calculated from the bottom to the tree root, each node may\nindicate its number of addresses. The end of the address assignment\nprocedure is marked by the definition of the ARTT at each node.\nA meshed tree can then be built on the top of an ART. This can\nbe done by adding additional links so that the network looks like\na mesh while each individual link perceives a tree as depicted in\nFigure 2.8.\nThe Meshed Adaptive Robust Tree (MART) allows routing a packet\nthrough a shorter path; single points of failure can be avoided. For\nN\nO\nK\nJ\nA\nB\nH\nI\nG\nE\nF\nD\nC\nL\nM\nFigure 2.8\nA Meshed Adaptive Robust Tree (MART).\n" }, { "page_number": 76, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n65\ninstance, if the link between H and B is broken, packets from H to\nC or to E can still be routed. However, paths are still non-optimal in\nmost cases.\nSamsung also proposes a key distribution scheme called KEYDS\nto provide security services. The mesh nodes that form the backbone\nshould provide security services to the rest of the network entities.\nEvery pair of backbone members shares a secret key that is used to\nsecure the communication between them. Besides, a group key is\nshared among all backbone members to allow backbone message\nbroadcasts. All mesh points should participate in the key pre-distri-\nbution scheme and should be able to perform common pair-wise\nkeys computations. The initial setup of the distribution key manage-\nment begins when each node within the mesh network obtains its\nID. Then, every mesh point obtains the key block from KEYDS with\na corresponding column of the incidence matrix. A member of the\nbackbone, as any other mesh point, also obtains the key block from\nKEYDS.\nIn addition, every member of the backbone obtains the corre-\nsponding key block from the trivial key pre-distribution scheme.\nThen, every mesh point (except members of the backbone) obtains\nthe final hash-value of the hash-chain and the lengths of the chain\nwith respect to that final value. Finally, every member of the back-\nbone obtains the start hash-value (the seed) of the hash-chain and\nthe current length of the chain with respect to the final value given\nto the mesh points. Key refresh decisions are then taken by the back-\nbone members when needed. When the network topology changes,\nthe key pre-distribution scheme executes the mesh point exclusion,\nthe mesh point association, and the lost mesh points’ recovery to\nadapt to the new network needs.\n2.4\nWireless Mesh LAN\nWireless mesh LANs have an extended coverage area compared to mesh\nPANs; they always adopt an infrastructure-based architecture and rely on\nreduced-mobility APs. Therefore, the PANs router mobility is no longer a\nchallenging issue. Nevertheless, mesh LANs need to provide QoS guaran-\ntees and address hand-off and roaming issues.\n2.4.1\nIntroduction and Advantages\nA wireless mesh LAN may be seen as a wireless LAN where all the APs\nare wirelessly interconnected. Traditional mobility management function-\nalities such as hand-over and roaming are supported; however, inter-AP\n" }, { "page_number": 77, "text": "66\n■\nSecurity in Wireless Mesh Networks\ncommunication within the same Extended Service Set (ESS) is done in a\nhop-by-hop fashion. The transmission scenario in a wireless mesh LAN is\ndone as follows: the AP managing the source forwards the traffic to its\nneighboring AP instead of sending it to all the APs in the ESS. Then, the\nneighboring APs sends the same packet to the next hop in the same way\nuntil the AP managing the destination is reached. At this time, the traffic is\nforwarded to the destination end node.\nIf we compare traditional wireless LANs to wireless mesh LANs, we\nnotice that the latter offers particular advantages related with the deploy-\nment costs, offered services, and nature of the supported applications. For\ninstance, deploying a mesh node needs no special wiring and configuration.\nWith little investment and easy configuration process, the network is more\nreliable because we can simply add as many wireless nodes as needed to\nincrease the performances and cover new zones. Mesh LANs also guaran-\ntee load-balancing and optimal resources utilization because wireless nodes\nmay act as routers or APs when the nearest AP is congested and route data\nto the closest low-traffic node. Fault tolerance is also provided because the\nclients communicate in a multi-hop fashion, exploiting the redundancy of\npaths in case of failures. The traffic is automatically rerouted while the failed\nrouters are rapidly detected and recovered or replaced. Furthermore, de-\nploying wireless mesh LANs addresses line-of-sight constraints, especially\nin outdoor environments. The provided applications in the mesh context\nfit particularly to the multi-hop architecture as explained as follows [16,17]:\n■\nWarehousing: Warehousing or broadband home networking appli-\ncations can be supported by traditional wireless LANs. However,\nthe APs are mainly installed on the roofs to provide good coverage;\nbesides, an expensive deployment of a wired backhaul is needed.\nAdopting wireless mesh LANs optimally addresses the pre-described\ndeployment issues. In fact, APs are wirelessly interconnected and\ncan be added anytime and anywhere to improve the scalability, the\nreliability, and the network performance. Moreover, fault-tolerant\npaths can be used to route the traffic between the mesh nodes until\nthe final destination while congestion resulting from the traditional\naccess to the hub is eliminated.\n■\nEnterprise networking: An enterprise local area network aims at\nsharing the enterprise resources while guaranteeing high transmis-\nsion rates and supporting advanced applications. It can be deployed\nin a small office, or it can interconnect multiple offices in the same\nbuilding or multiple offices in separate sites. Traditional wireless\nLANs have been widely adopted to reduce the internetworking costs\nwhile improving the scalability. Nevertheless, the need of deploying\na wired infrastructure has been always present. Moreover, adding\nnew APs to the backhaul locally enhances the WLAN capacity, but\n" }, { "page_number": 78, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n67\ndoes not guarantee the fault tolerance and the congestion reduc-\ntion. Adopting wireless mesh LAN architecture enables the share of\nresources and an overall performance enhancement, thanks to the\nmulti-hop communication and the wireless infrastructure deploy-\nment. In fact, bottleneck congestion resulting from the one-hop\naccess to the traditional APs is eliminated. Besides, the infrastructure\ncan easily scale according to the network’s needs without requiring\ncomplex configurations and wiring.\n■\nHealthcare: The hospitals are always built to prevent the propagation\nof electromagnetic waves because any disruption can have catas-\ntrophic consequences. However, exchanging voluminous monitoring\nand diagnosis data such as high-resolution radiographs at real-time\nand sharing information between the hospital crew is becoming a\npressing need. The deployment of a wired network only intercon-\nnects some fixed medical devices while the adoption of a traditional\nwireless LAN induces high backhaul-wiring costs and many dead\nzones. The optimal solution consists of deploying a wireless mesh\nLAN where the mesh nodes and routers are placed according to\npropagation characteristics and capacity needs.\n2.4.2\nArchitecture Technologies\nThe mesh wireless LAN has two possible architectures. The infrastructure\narchitecture is formed by different APs interconnected wirelessly within\nan ad hoc network. The resulting wireless backhaul reacts to any topology\nchanges by processing automatic topology learning and dynamic path con-\nfiguration. The IEEE 802.11s standard defines the physical and MAC func-\ntions needed by the interconnected APs to manage the mesh clients such\nas the reliable unicast or multicast/broadcast delivery. The infrastructure\narchitecture aims at reducing deployment costs while enhancing network\ncoverage and reliability. More specifically, it becomes easy to add new APs\nto enforce the existing backhaul network and cover dead zones without any\nneed of wire deployment and complex configurations. The infrastructure\nmeshing is the most used because it allows good scalability and supports\ngateway functions such as bridging, thus enabling the connection to the\nInternet and the integration with other network technologies.\nThe client meshing architecture does not require the backhaul; in fact,\nmesh nodes can play the role of APs and be clients and routers at the\nsame time forming a dynamic ad hoc network. To do so, the mesh nodes\ncommunicate in a peer-to-peer fashion and perform layer 3 routing while\nsupporting auto-configuration and providing end user services. Packets are\ntransmitted within flat network architecture from one hop to another until\nthe final destination; however, congestion occurs more frequently and the\n" }, { "page_number": 79, "text": "68\n■\nSecurity in Wireless Mesh Networks\nnetwork performance rapidly decreases when the number of mobile nodes\ngrows. The hybrid architecture combines the infrastructure and the client\nmeshing to achieve enhanced performances. Mesh clients can be managed\nby APs, but may also directly communicate with other peers. This mode is\nstill not used very often in case of WiFi meshes.\n2.4.3\nChallenges\nA wireless LAN implementing the IEEE 802.11 standards is formed by one or\nmore APs responsible for central management and a set of mobile stations\nequipped with a 802.11-compliant interface. An AP and the stations situated\nin its coverage zone form a cell or Basic Service Set (BSS). The mobile\nstations may also form an Independent Basic Service Set (IBSS) when they\ndirectly communicate in an ad hoc fashion without requiring a central AP.\nA set of APs may be interconnected by a wired distribution system, thus\nforming an ESS which can be viewed as a single 802.11 network segment.\nIn the mesh context, the meshing APs have to form a wireless infra-\nstructure; therefore, they need to implement auto-configuring mechanisms\nto automatically integrate the ad hoc network formed by the neighbor-\ning APs. Besides, the mesh traffic originated by a node is handled by the\nmanaging AP which is responsible for its delivery to the destination. This\ntraffic may cross multiple intermediate nodes before reaching the recipient\nand each crossed node will introduce some latency, thus hardening the\nQoS provision in terms of minimum delay and jitter. Meanwhile, APs need\nto exchange data on wireless channels; therefore, mesh networks should\nguarantee the coexistence of intra-BSS and inter-BSS communication by\neliminating possible interference while guaranteeing the required QoS [16].\nHidden and exposed terminals problems should also be addressed. Last\nbut not least, APs forward the arriving packets to their MAC layer, which\nadopts a drop-tail queue management without taking into consideration the\nnumber of crossed hops. This management strategy may lead to a severe\nunfairness problem because neighboring or smaller hop length flows arrive\nmore frequently at APs and fill up the link layer buffer. Consequently,\npackets coming from far away nodes face a full buffer and will systemati-\ncally be dropped.\n2.4.4\nThe IEEE 802.11s Standard\nAs described so far, the IEEE 802.11 standards define physical and MAC\nmechanisms for one-hop communications, rely on a wired infrastructure,\nand are subject to throughput degradation and unfairness when applied to\nmulti-hop communication scenarios. Being aware of the tremendous advan-\ntages offered by mesh networks, industrial actors and researchers formed\n" }, { "page_number": 80, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n69\na separate task group in May 2004 under the 802.11 Working Group called\nIEEE 802.11s ESS Mesh that aims at specifying the physical and MAC exten-\nsions needed for the multi-channel support. Two main proposals, denoted\nby SEEMesh and Wi-Mesh, merged in January 2006 and were confirmed\nunanimously in March 2006. This fusion has resulted in the embryo of the\n802.11s standard that will probably be approved in 2008.\nThe 802.11s standard aims at specifying the architecture and protocols\nrequired for the implementation of a Wireless Distribution System (WDS).\nThe mesh mobile nodes will process an automatic self-configuration as\nsoon as they enter the mesh network while the routing protocol will be\nintegrated in the MAC layer to allow a dynamic path configuration for\nbroadcast/multicast and unicast traffic. When the mesh traffic should reach\na destination which is not associated with the AP of the sender, the AP\nwill not send the packets to all APs within its ESS as in IEEE 802.11; it\nwill rather send them to the next AP on the path. The mobile devices\nwill support the multi-channel communications and can be equipped with\nmultiple radios using the same mode while the targeted frequency band\nwill be the unlicensed 2.4 to 5 GHz to guarantee the interoperability with\nother 802.11 standards.\n2.4.4.1\nIEEE 802.11s Device Classes\nThe 802.11s architecture is based on different classes of devices, as illus-\ntrated in Figure 2.9. A Mesh Point (MP) may be an AP or a mobile station\nwhich provides a partial or full mesh relaying function. An MP processes\nneighbor discovery and selects the channel to communicate and forward\nMAP\nLegacy 802.11s links\nMP\nMP\nMPP\n802.11s mesh links\nSTAs\nMP\nFigure 2.9\nThe proposed 802.11s architecture.\n" }, { "page_number": 81, "text": "70\n■\nSecurity in Wireless Mesh Networks\nthe traffic for other MPs using bidirectional channels. Mobile stations or end-\nuser devices or stations are traditional stations with no mesh capabilities.\nSuch devices will be wirelessly interconnected to a mesh AP (MAP) which\nis a particular MP able to operate in one of the legacy 802.11 modes. The\n802.11s standard defines the mesh portal (MPP) that interconnects multi-\nple WLAN meshes. The MPP can also play the role of an entry or exit\nto a wired network and support advanced functions such as transparent\nbridging, address learning, layer 3 routing, and bridge-to-bridge commu-\nnications. Finally, an MPP may be configured for topology building and\nelected to become the root of the default forwarding tree, thus becoming\na root portal. Each mesh network is identified by a mesh ID which is the\nequivalent of a service set identifier (SSID) representing an ESS in legacy\n802.11 networks.\n2.4.4.2\nMedium Access Control: The Medium Access\nCoordination Function\nThe Medium-access Coordination Function (MCF) is a MAC sub-layer which\nis built on the top of the physical layer to provide the mesh services. As de-\npicted in Figure 2.10, the MCF is responsible for guaranteeing the mesh con-\nfiguration and management, the mesh security services based on the 802.11i\nstandard, the topology discovery and association, the topology learning,\nMesh security\nDiscovery and\nassociation\nMedium access\ncoordination\nMesh topology\nlearning routing and\nforwarding\nMesh interworking with other 802 networks\nMesh\nmeasurement\nPHY5\n802.11 service\nintegration\nMesh configuration and management\nFigure 2.10\nThe 802.11s MCF function.\n" }, { "page_number": 82, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n71\nthe routing and forwarding functions, the medium access coordination,\nthe mesh measurement, and the mesh internetworking with other IEEE 802\nnetworks.\n■\nMesh Topology Learning, Routing, and Forwarding: The mesh topol-\nogy learning and forwarding function is processed by the MP to\ndiscover its neighbors. It allows automatic topology learning and\nenables the link establishments and the dynamic paths discovery\nfor data delivery purposes.\nWhen a new MP enters the mesh network, it begins by collecting\ninformation from neighboring MPs either by sending a probe request\nor passively listening to the periodic beacons. The candidate MP\ncan then choose to associate with another peer to form the mesh\ntopology. This association highly depends on the peer’s capability,\nits power, its security information, and its link quality.\n■\nPath Selection Protocol: The MCF sub-layer implements the rout-\ning function at the MAC. In fact a hybrid routing protocol sup-\nporting both fixed and mobile MPs and including proactive and\nreactive schemes should be defined to handle unicast and multi-\ncast/broadcast traffic delivery. The 802.11s Standard Committee has\nchosen to mix the Ad-hoc On-demand Distance Vector (AODV, [51])\nand the Optimized Link State Routing (OLSR) protocols while defin-\ning a set of radio-aware metrics reflecting the link status to enhance\nthe routing reliability. For instance, an airtime metric reflecting of\nchannel, path, and packet error rate has been proposed in [57] while\nthe WRALA metric (Weighted Radio and Load Aware [19]) reflects the\nprotocol overhead at the MAC and PHY layers, size of the frame, bit\nrate, link load, and error rate.\n■\nForwarding Scheme: The wireless LAN mesh network uses four-\naddress data frames with two extensions for QoS support and mesh\ncontrol, as depicted in Figure 2.11. Each MP which receives a data\nframe begins by checking its authenticity and destination MAC and\nthen forwards it if everything is OK. As STAs transmit three-address\nframes, the correspondent MPA needs to convert them to the four-\naddress format before forwarding them toward the destination. Multi-\ncast and broadcast traffic is also forwarded if it uses the four-address\nformat; moreover, the time to live (TTL) sub-field is decremented by\neach intermediate MP to monitor the broadcast data in the WLAN\nmesh.\n■\nMedium Access Coordination: The Medium Access Coordination\nsub-layer that has been proposed in [57,58] implements the enhan-\nced distributed channel access (EDCA) mechanism used in 802.11e\n[20]. This sub-layer also provides congestion control, power saving,\nsynchronization, and beacon collision avoidance. Multiple channel\n" }, { "page_number": 83, "text": "72\n■\nSecurity in Wireless Mesh Networks\nMesh\nforwarding\ncontrol\nQoS\ncontrol\nAddr\n4\nSeq\ncontrol\nAddr\n3\nAddr\n2\nAddr\n1\nFrame\ncontrol\nDur\n0−2312\nBody\n4\nFCS\nMAC header\n802.11e QoS header\nBytes 2\nMesh E2E seq\nLoop free\nKeep receive order &\navoid duplication\nTime to live\n2\n2\n2\n3\n6\n6\n6\n6\nFigure 2.11\nThe 802.11s mesh data frame.\noperations which are based on the common channel framework\n(CCF) [21] are also supported in multiradio, single radio, or hybrid\nenvironments.\n■\nMesh Configuration and Management: Mesh networks rely on node\nself-configuration to accelerate and facilitate the deployment. There-\nfore, mesh nodes need to implement automatic management mod-\nules and association protocols that enable the MPs associating with\nother MPs neighbors and even external nodes. Management func-\ntions should be able to detect the failed nodes to replace them\nalthough the mesh network is to a certain extent failure-tolerant.\nThe format of a management frame is shown by Figure 2.12; it\nincludes the DA (destination address) or receiving MP MAC address,\nthe SA (source address) or transmitting MP MAC address, and the\nBSSID (basic service set ID) field stating for the wildcard value.\nIt is worth noticing that the interfaces need to implement the\n802.11h to enable compliance with dynamic frequency selection\nBytes 2\n4\n0−2312\nFrame\ncontrol\nBSSID\nDuration\nDA\nSA\nSeq\ncontrol\nFrame\nbody\nFCS\nWildcard value\nMPs’ addresses\nMac header\n2\n6\n6\n6\n2\nFigure 2.12\nThe 802.11s mesh management frame.\n" }, { "page_number": 84, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n73\n(DFS) requirements and enhance the efficiency of the multi-hop\ntransmissions, the power saving, and the total capacity.\n2.4.5\nRouting and QoS Support\nUsing the mesh network architecture allows a wide coverage, thanks to\nmulti-hop ad hoc communication, but requires a particular QoS manage-\nment, especially because the mesh nodes act as routers and clients at the\nsame time and do not rely necessarily on a centralized management point. A\nQoS routing protocol has been proposed for a wireless mesh LAN infrastruc-\nture called WMR [21] that supports multimedia applications by guaranteeing\nminimum bandwidth and maximum end-to-end delay for all intra-BSS and\ninter-BSS communications.\n2.4.5.1\nWMR Protocol Overview\nThe WMR protocol is based on the Ad hoc QoS Routing (AQOR) protocol\nthat has been developed for the MANET context by the authors in [60].\nIt is based on the following phases: topology discovery, route discovery,\nadmission control with QoS constraints, and route recovery.\n■\nTopology Discovery: The topology discovery phase consists of\nexchanging local information with the mesh nodes neighbors to get\nan updated view of the current topology and estimate the distance\nto the backhaul. Each mesh node maintains a distance Tag D(I) that\nindicates the number of hops to the nearest AP; it is set to 0 for APs\nand to 16 for each newcomer. Moreover, each mesh client and AP\nwithin the network should periodically send a Hello message with\nTTL field set to 1 and a tag field indicating the distance to the nearest\nAP. This control message is then used to update a list of neighbors\nN[I] and determine the distance from the nearby AP.\n■\nRoute Discovery: The route discovery is processed on-demand by\nsending a Route Request for route exploration and then waiting for\nthe correspondent Route Reply enabling the route registration. The\ntraffic addressed to nodes that do not belong to the mesh network\nis sent to the nearest AP as if it was the final destination.\n■\nRoute Exploration: Each node wishing to communicate has to\nsend a Route Request while indicating its QoS requirements in\nterms of minimal bandwidth and end-to-end delays. The route\nexploration algorithm differs according to the nature of the des-\ntination node. In fact, if the destination is internal to the mesh\nnetwork, the Route Request is assigned a TTL value and then\nflooded. However, if the traffic is addressed to an external node\n(e.g., a node that does not belong to the mesh network such\nas an Internet destination), the chosen multi-hop wireless path\n" }, { "page_number": 85, "text": "74\n■\nSecurity in Wireless Mesh Networks\nto the AP should be as short as possible to guarantee good\nroute stability and channel efficiency. Therefore, a distance-\nconstrained discovery algorithm based on the distance tag infor-\nmation stored at every crossed node is proposed. In fact, the\nsource includes its distance tag in the request, and then only the\nnodes having a smaller value should receive the control packet,\nupdate it by setting their own distance tag, and forward it. An\ninitial sequence number equal to zero is set for each Route Re-\nquest and updated so that only the first accepted packet of a flow\nis relayed during one round of the control packet propagation,\nthus minimizing the overhead and reducing the traffic aggre-\ngation induced by the multi-hop flow. When a node receives\nthe Route Request, it checks whether its available bandwidth is\nequal or superior to the required one. If it is the case, the flow\nis accepted, a new entry is added to the routing table with the\nstatus explored, and the packet is forwarded.\n■\nRoute Registration: The destination node should send a Route\nReply on the reverse path to the source for every received Route\nRequest. When receiving the reply, intermediate nodes re-\nestimate their available bandwidth and update the routing table\nentry by setting the status registered, but the effective bandwidth\nreservation is only done after receiving the first data packet. All\nintermediate nodes of all established paths will still be in the\nregistered status for a period of 2 * Tmax, where Tmax is the\nmaximum end-to-end delay of the requesting flow. If no data\npacket of the correspondent flow arrives within the threshold\nperiod, the route will be released.\n■\nAdmission Control: The admission control decision is performed at\nevery node during the exploration phase to discover paths. There-\nafter, the route offering the shortest end-to-end delay will be chosen\namong the paths providing the minimal requested bandwidth.\n■\nBandwidth Control: To estimate whether a flow can be trans-\nmitted over a path while providing the bandwidth-specified re-\nquirements, a correct estimation of the available link capacity\nand the truly consumed bandwidth is required. As wireless links\nare shared among all neighboring nodes, the available band-\nwidth at a node I is determined by the raw data rate of that\nnode and the neighboring transmissions. This available band-\nwidth value is continuously changing due to the node’s mobility.\nBesides, the bandwidth consumed by a flow (j) is different from\nthe minimal bandwidth required by that flow due to the inter-\nference caused by neighbors. To estimate bandwidth values, a\n" }, { "page_number": 86, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n75\nhalf duplex channel and identical data rates and transmission\nrange for all nodes have been assumed [21]. The available band-\nwidth at a node I is estimated by computing the existing total\nchannel traffic load, which includes the traffic generated by I\nand its neighbors, I’s neighboring traffic, and finally the bound-\nary traffic crossing the boundaries of I’s range and exchanged\nby I’s neighbors and nodes that are outside I’s range. Finally,\nto estimate the bandwidth that should be reserved for a flow\n(j), both the new self traffic and boundary traffic introduced by\nthe requesting flow were considered [21]. After computing the\navailable bandwidth and the required minimum bandwidth, the\nadmission control compares these results to determine whether\nto accept the flow.\n■\nEnd-to-End Delay Control: A proposal was put forth in [21] to\nestimate the delay from the source to the destination denoted by\nTup and the delay back to the source Tdown and verify whether\nTup + Tdown < 2Tmax, where Tmax is the maximum tolerated\ndelay. Because many paths may be found, the route on which\nthe route reply arrives first is chosen. If no reply arrives within\n2 T max, the source may later retry the route discovery or turn\ndown the flow.\n■\nRoute Recovery: Discovered routes may be broken due to node\nmobility or channel deterioration, thus leading to QoS violations.\nTo address this issue, the destination node estimates the end-to-\nend delay experienced by the arriving data packets and triggers the\nQoS recovery mechanism when needed. With a traditional ad hoc\nrouting algorithm, an intermediate node that does not receive the\nhello packet from its neighbor after a time-out notifies the source by\nsending an error packet. Consequently, the path problems cannot\nbe detected at real-time and resolved quickly. WMR detects a QoS\nviolation using the bandwidth reservation information at the destina-\ntion node. In fact, the destination triggers the recovery mechanism\nwhen it does not receive the data packets before the reservation\ntime-out. Besides, an intermediate node may send an error notifica-\ntion back to the source if the next hop cannot be reached to release\nthe reserved resources.\n■\nSimulation Results: The WMR [21] protocol simulation has been done\nusing OPNET Modeler 7.0, which was modified to support multi-\nhop communications. The MAC layer module was the default IEEE\n802.11 DCF and the WMR was inserted on top of it. The authors have\nalso supposed that all nodes had a transmission range of 200 meters\nand a raw bandwidth of 2 Mbps. The maximum packet size used\n" }, { "page_number": 87, "text": "76\n■\nSecurity in Wireless Mesh Networks\nin temporary bandwidth reservation was set to 1024 bytes while the\nsender buffer was set to 64 packets. A source node might retry the\nroute discovery three successive times. Hello messages were sent\nevery second and the neighbor time-out was set to three seconds.\nForty nodes were randomly deployed in a 800 m * 800 m range and\nten flows were randomly spread among these nodes; the network\nalso included two APs located at diagonal corners of the field. The\nsimulation period was set to 300 seconds. Stream media applications\nused Constant Bit Rate (CBR) flows with ten packets per second and\nfixed data packet size of 1024 bytes. All flows tolerated a maximum\ndelay Tmax equal to 0.1 second and required a minimum bandwidth\nof 80 kbps. The performance metrics that have been considered\nwere (1) the traffic admission ratio, (2) the end-to-end delivery ratio,\n(3) the average end-to-end delay, (4) the ratio of late packets, and\n(5) the normalized routing overhead.\nThe traffic admission ratio is the ratio between the number of\ndata packets sent to the network from the sources and the number\nof data packets generated at the sources up to time T. The end-to-\nend delivery ratio is the ratio between the number of data packets\nthat arrive at the destination and the number of data packets sent\nfrom the source up to time T. The average end-to-end delay is the\naverage end-to-end delay of data packets received at the destina-\ntion up to time T, including all possible delays caused by buffering\nduring route discovery, queuing delay at the transmission queue,\nretransmission delays at the MAC, and propagation delay. The ratio\nof late packets is the ratio between the number of data packets that\nexceed the delay bound and the number of data packets that arrive\nat the destination up to time T. Finally, the normalized routing over-\nhead is the number of control packets transmitted per data packet\narrived at the destination up to time T.\nThe simulation results showed that WMR has succeeded in pro-\nviding the required QoS while adapting to the network changes and\nminimizing the control overhead [21]. Nevertheless, we believe that\nWMR provides QoS within the mesh network; that is, when mesh\nnodes communicate with external ones (e.g., Internet nodes), the\nQoS is only provided on the sub-path between the source and the\nAP. We think that the AP needs to perform a re-estimation of the\nrequired QoS in terms of minimum delay by taking into consider-\nation the time already spent when crossing the intermediate mesh\nnodes Tcross until the AP. It is clear that if the minimum delay is\nclose to Tcross, it will be difficult to provide the required end-to-end\nQoS. Finally, the WMR did not provide an optimal mechanism for\neffectively achieving routes recovery in case of paths breakage.\n" }, { "page_number": 88, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n77\n2.4.6\nOverview of Available Commercial Systems\n■\nStrix Systems: The Access/One® Network powered by Strix Sys-\ntems provides a wireless LAN system that supports multiple radio\nfrequency technology within a scalable network [14]. Access/One\nNetwork wireless APs deployed within a mesh architecture can auto-\nmatically discover their neighbors and route traffic choosing optimal\npaths according to environment conditions changes. For this pur-\npose, each node identifies the optimal route to the closest and least-\ncongested network server (an Access/One Network module used for\ncontrol signaling and data registry) and a path to the wired links via\nmesh nodes. When new nodes integrate the network or congestion\noccurs on the wireless links, the established routes are automatically\nre-evaluated to guarantee the maximum performances. Moreover,\nthe network modules scan all available channels in real-time to\ndefine a list of potential reachable client modules. Particular radios\nmay be dedicated for particular functionalities (either send or\nreceive) and the least-congested channels are selected to build the\nmesh. Furthermore, Access/One Network nodes guarantee the au-\nthentication by supporting encapsulated RADIUS exchanges, includ-\ning the MD5, TLS, TTLS, and PEAP mechanisms. Besides, privacy is\nprovided using the supported WEP, including TKIP/MIC enhance-\nments, and AES cipher suites, with either static or dynamic keys.\nFinally, Access/One Network nodes support the IEEE 802.1q VLAN\ntagging of wireless frames and assign priorities to them so that they\ncan be processed by a VLAN-aware switch.\n■\nTropos® Networks: Tropos Networks propose the MetroMeshTM\nNetworks architecture that provides WiFi clients with a secure access\nto network services in a coverage area ranging from local to metro-\npolitan [15]. For instance, the Tropos 3210 indoor MetroMesh router\nimplements the proprietary Predictive Wireless Routing Protocol\n(PWRP) to create a self-organizing and self-healing wireless mesh by\nsearching for the optimal data path to the wired network. The Tro-\npos 3210 indoor MetroMesh router guarantees wireless connectivity\nto standard 802.11b/g clients. Moreover, it seamlessly meshes with\nthe Tropos 5210 outdoor MetroMesh router to extend the coverage\narea of the metro-scale WiFi network. The supported MetroMesh OS\nprovides the VLAN technology and implements the auto-discovery\nand auto-configuration on power-up with a real-time adjustment of\nthe established paths to guarantee optimal performances. Secure\nmanagement features include AES encryption of wireless routing,\nMAC address access control lists definition, and a full VPN compat-\nibility. Thanks to such mechanisms, individual users with different\n" }, { "page_number": 89, "text": "78\n■\nSecurity in Wireless Mesh Networks\nprivileges and security needs may operate independently while max-\nimizing network economics and performance.\n2.5\nWireless Mesh MAN\n2.5.1\nPurpose\nComplex multimedia applications are becoming very popular, leading cus-\ntomers to request the marriage of mobility support with a high bandwidth\nand an enhanced availability, reliability, and flexibility. As cellular-based\ntechnologies have not been satisfactory in many aspects, broadband wire-\nless access is gaining the interest of researchers and network operators\nwhile multi-hop communication is expected to become the leading technol-\nogy. The aim of the mesh metropolitan networks is to provide broadband\naccess everywhere and anytime by increasing reach and coverage through\nmultiple hops, without compromising performance or reliability. Some of\nthe IEEE 802.16 standards have provided the mesh network support and\ntried to minimize the impact of multipath interference while providing con-\nnectivity between network endpoints without direct line-of-sight.\n2.5.2\nTargeted Services\nCompared to wired or cellular networks, wireless mesh MANs are an eco-\nnomic alternative to enable ubiquitous broadband networking with high\nthroughput and multimedia-applications support even for underdeveloped\nregions. Targeted services are mainly wireless Internet access, public safety,\nand implementation of intelligent transportation systems.\n■\nISP: Internet service providers are searching for integrated solutions\nthat provide public Internet access for residents, enterprises, and\ntravelers with consistent levels of service and pricing, guaranteed\nscalability, and minimal investments. On the other hand, countries\nand cities are encouraging the deployment of information technolo-\ngies to improve government services which will attract business and\ncitizens and boost the economic development. A growing number\nof ISPs have found in the wireless mesh networks an ideal solution\nto provide both indoor and outdoor broadband wireless connectiv-\nity in urban and rural environments without the need for costly\nnetwork infrastructure. With a Wireless Internet Service Provider\n(WISP), users are able to connect to the Internet when they travel\noutside their home or business, or go to another city that also has\na WISP. As examples, the city of Chaska, Minnesota, has formed\nchaska.net, a WISP that provides low-cost, high-speed Internet\n" }, { "page_number": 90, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n79\nconnections to more than 7,500 homes and 18,000 residents [22].\nThe city of Moorhead, Minnesota, has also succeeded in installing a\nmetro-scale broadband WiFi network from Tropos Networks, which\nprovides lower-cost Internet access anywhere in the city [23].\n■\nPublic Safety: Municipal police, fire, and emergency departments\nhave a pressing need for adopting metro-scale mesh networks and\nthe resulting mobile broadband data access. In fact, public safety\nagents have used mobile data radio systems for years, but the im-\nplemented cellular networks offered near-ubiquitous coverage and\nlow data rates (9.6 kbps), thus prohibiting in-field access to multi-\nmedia data and applications. Adopting metro-scale mesh networks\nfor mobile broadband data access will improve the effectiveness and\nefficiency of public safety officers by getting critical information in\ntheir hands on the street in a totally secure manner. Furthermore,\ndeploying metro-scale video surveillance (e.g., in high crime areas\nand strategic targets) will enhance public safety and bring appli-\ncations such as virtual lineups, fingerprint analysis, and access to\ndetailed mug shots or floor plans out of the station house and into\nthe field where they are needed. Besides, equipping firemen with\nlocator chips and helmet-mounted wireless video cameras can help\nincident commanders and field crews share knowledge during emer-\ngencies.\n■\nIntelligent Transportation Systems: Mesh networking technology can\nbe adopted by transportation companies to provide intelligent trans-\nport systems, if a high-speed mobile backhaul from a vehicle to the\nInternet is supported. Buses, ferries, and trains equipped with wire-\nless mesh access can provide real-time travel information, allow re-\nmote monitoring of in-vehicle security video, permit the addressing\nof transportation congestion, and help control the pollution.\n2.5.3\nArchitecture\nBroadband wireless MAN standards detail two modes of communication:\nthe Point-to-Multipoint (PMP) mode and the mesh mode. With the PMP\nmode, the subscriber station (SS) can only communicate with a base station\n(BS) using separate downlink and uplink sub-frames [28]. Consequently, the\nBS always has to route data between two communicating SSs [29]. The mesh\nmode adopts a multi-hop communication by allowing every station (sub-\nscriber or base station) to directly communicate with other stations in the\nnetwork, independently of their nature. Thus, traffic can be routed through\nother SSs and occur directly between SSs while the mesh BS connects the\nwireless network to the backhaul links. An adaptive scheduling mechanism\nis used to allocate mini slots and associated channels within the data sub-\nframe. The assignment of transmission opportunities in the direct links can\n" }, { "page_number": 91, "text": "80\n■\nSecurity in Wireless Mesh Networks\nbe controlled by either a centralized or distributed algorithm; furthermore,\na three-way handshake is always used to request, grant, and confirm those\ntransmission opportunities.\n■\nCentralized Scheduling: In centralized scheduling, the BS has to pro-\nvide the schedule configuration for the SSs within a threshold num-\nber of hops after analyzing the transmission requests. Consequently,\nthe BS has the same functionality as in the PMP mode. However, not\nall the SSs have to be directly connected to the BS because some\nof them can determine the actual schedule for their direct neigh-\nbors from these flow assignments [61]. The centralized scheduling is\ncoordinated because the scheduling packets are transmitted within\nscheduling control sub-frames without risks of collision. It is partic-\nularly adapted for the transmission of persistent traffic streams.\n■\nDistributed Scheduling: In distributed scheduling, the mesh BS does\nnot coordinate the process in a centralized manner. In fact, all sta-\ntions (BS and SS) have to coordinate their transmissions with their\ntwo-hop neighbors and broadcast their schedules to all their direct\nneighbors. Each request is analyzed by the granter using a given\nslot allocation algorithm; then the granter returns a grant message\nin case of success. In this case, the requester sends back the received\nmessage to acknowledge its reception. The distributed scheduling\nmay be coordinated or uncoordinated. The coordinated distributed\nscheduling uses the scheduling packets transmitted within the con-\ntrol sub-frame. The uncoordinated distributed scheduling fits to oc-\ncasional or brief traffic over links which have not been considered\nby the current centralized or coordinated distributed schedule. It is\nperformed in a contention-based manner where scheduling control\nmessages are sent during the data sub-frame while avoiding con-\nflict with the schedules already established using the coordinated\nprocedures [40].\n2.5.4\nStandards\nThe IEEE 802.16 standards, also known as WiMAX (Worldwide Interoper-\nability for Microwave Access), is currently viewed as the future technology\nthat will be adopted for the deployment of broadband wireless metropolitan\narea networks [28]. The physical layer detailed by the IEEE 802.16 standards\nuses the frequency ranges 2 to 11 GHz and 10 to 66 GHz and supports\nsingle carrier (SC), Orthogonal Frequency Division Multiplexing (OFDM)\nand Orthogonal Frequency Division Multiple Access (OFDMA). The 2 to\n11 GHz has no line-of-sight requirements; however, it induces multi-path\n" }, { "page_number": 92, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n81\nControl\nsubframe\nControl\nsubframe\nData subframe\nData subframe\nTransmission\nopportunity\nTransmission\nopportunity\nFrame n\nFrame n + 1\nMinislot\nMinislot\nMinislot\nFigure 2.13\n802.16 MAC frame in mesh mode.\nand requires additional functionalities such as power management, error\nrecovery, and interference mitigation. The MAC layer which manages the\nshare of the common channel resources adopts the Time Division Multiple\nAccess (TDMA) and supports both PMP mode and mesh mode. In the fol-\nlowing section, we detail the PHY and MAC extensions needed to support\nmesh mode.\n2.5.4.1\nMAC Layer Overview in WiMAX Mesh Mode\nThe mesh mode defined by the IEEE 802.16 standard supports only Time\nDivision Duplex (TDD), which separates uplink and downlink in time. A\nMAC frame in mesh mode is made up of two sub-frames fixed in length,\nthe control sub-frame and the data sub-frame, as illustrated by Figure 2.13.\nThe data sub-frame illustrated by Figure 2.14 is used for data transmis-\nsion in a link connection-oriented basis (there is no end-to-end connection\n[42]). One link is used for bidirectional data transfers between two SSs\nwithout distinction between uplink and downlink sub-frames (per-analysis\nmesh mode).\nPhysical bursts vary in length; they are made up of a preamble followed\nby MAC PDUs. The latter includes a fixed-length MAC header, a fixed-length\nmesh sub-header, a variable length payload, and an optional CRC field. The\ncontrol sub-frame is only used for the signaling message transmission trans-\nfers. It serves the cohesion, creation, and maintenance between all SSs and\nto the data scheduling [41]. The parameter MSH CTRL LEN determines the\nnumber of transmission opportunities that can be carried by one control\nsub-frame, and ranges between 0 and 15. Besides, each transmission op-\nportunity has the length of 7 OFDM symbols. Consequently, the total length\nof a control sub-frame is computed by Lcs = 7 ∗MSH CTRL LEN. A con-\ntrol sub-frame can be a network-control sub-frame or a schedule-control\nsub-frame, as illustrated by Figure 2.15.\nThe network control sub-frame is useful for new terminals that want to\naccess the network because it is used to advertise network information and\n" }, { "page_number": 93, "text": "82\n■\nSecurity in Wireless Mesh Networks\nPHY burst\nSS # j\nPHY burst\nSS # k\nLong\npreambule\nMAC\nPDU # 1\nMAC\nPDU # m\nPadding\nMAC\nheader\nMAC payload\nCRC\n6 bytes\n0−2039\nTDM portion\nMesh\nsub-header\n2\n4\nFigure 2.14\n802.16 data sub-frame in mesh mode.\nsynchronization elements [34]. In fact, active nodes periodically broadcast\nthe MSH-NCFG message containing basic configuration information such\nas the BS identifier and the base channel in current use [35]. A new node\nthat wants to access the mesh starts listening to the MSH-NCFG to pinpoint\nactive networks. Based on the advertised information, it establishes a coarse\nsynchronization and starts the network entry process.\nThe network entry process begins when a joining node, also called a\ncandidate node, selects one sponsoring node and sends the network entry\nM MSH−DSCH scheduling \ncontrol transmit opportunities \nMSH−NENT \ntransmit opportunity \nN−1 MSH−NCFG \ntransmit opportunities \nCentralized scheduling \ncontrol (MSH−CSCH & \nMSH−CSCF) minislots \nData sub-frame\nScheduling\ncontrol sub-frame\nFrame, addressed by a\n12−bit frame number\ndivided into (up to) 256 minislots\nNetwork\ncontrol sub-frame\nFigure 2.15\nThe MAC control sub-frame in mesh mode.\n" }, { "page_number": 94, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n83\nCentralized scheduling\nLong\npreambule\nMSH−CSCF or\nMSH−CSCH\nGuard\nsymbol\n2 symbols\n1 symbol\nDistributed scheduling\nLong\npreambule\nGuard\nsymbol\n2 symbols\n1 symbol\nMSH−DSCH\nFigure 2.16\nThe schedule control sub-frame in mesh mode.\nmessage MSH-NENT:Request, including provider configuration data and\noptional authentication code. The sponsoring node responds by the MSH-\nNCFG:NetEntryOpen message advertising the candidate’s MAC address as\nbeing sponsored and including initial schedule. The new node acknowl-\nedges by sending a MSH-NENT:Ack; then higher-layer DHCP configuration\nand authentication are processed. Finally, the new node sends the MSH-\nNENT:Close and the sponsor responds with the MSH-NCFG:Ack [40]. If the\nselected sponsor does not advertise the new node’s MAC address, then the\nprocedure is repeated MSH-SPONSOR-ATTEMPTS times using a random\nback-off between attempts. A new sponsor is selected when all attempts\nfail.\nTo request bandwidth, SSs send connection-based requests in stand-\nalone or piggyback messages, including required numbers of bytes. Band-\nwidth is then allocated on an SS basis. The schedule control sub-frame\ncarries the scheduling information of the data sub-frame transmission op-\nportunities. It is also divided into two parts: the centralized scheduling\nmechanism (CSCH) and the distributed scheduling mechanism (DSCH),\nas detailed in Figure 2.16. When centralized scheduling is adopted, the\nmesh BS periodically collects network information and resources reserva-\ntion demands while the SS sends its resource allocation request to the BS\nencapsulated in a CSCH:Request message. The corresponding CSCH:Grant\nis created by the BS and broadcasted to the SSs within a threshold hop\nrange; then those SSs shall forward the received message to their neigh-\nbors that are further away from the BS (i.e., more hops to the BS). The\nCSCH includes the following parameters [31]:\n■\nFlow Scale: Determines scale of the granted bandwidth\n■\nNumAssignments: Number of 8-bit assignment fields followed\n■\nUpstreamAssignment: Base of the granted bandwidth as bits per\nsecond for the ingress traffic of the node in the BS routing tree\n■\nDownstreamAssignment: Base of the granted bandwidth as bits per\nsecond for the egress traffic of the node in the BS routing tree\n" }, { "page_number": 95, "text": "84\n■\nSecurity in Wireless Mesh Networks\nWhen distributed scheduling is adopted, request and grant of channel\nresource are delivered by an MSH-DSCH message among nodes.\nIn coordinated distributed scheduling, all the stations (BS and SS) peri-\nodically transmit the MSH-DSCH in a collision-free fashion to inform neigh-\nbors with the schedule of transmissions. The mesh distributed election-\nbased scheduling used for scheduling the MSH-NCFG and the coordinated\nMSH-DSCH control messages guarantees collision-free scheduling within\neach node’s extended neighborhood. The algorithm is run when the local\nnode should transmit (NextXmtTime = now); its inputs are as follows:\n■\nThe frame number and the transmit opportunity number within that\nframe for the type of message being scheduled\n■\nAll the node’s identifiers within the two or three hops neighborhood\n■\nThe XmtHoldoff Time of the local node, which is the node transmit\nhold-off delay\n■\nAs many couples of {node ID, NextXmtTime, XmtHoldoffTime} of\nnodes within the two or three hops neighborhood as have been re-\ncently received, where NextXmtTime is the node’s next transmission\ntime of MSH-NCFG\nThe algorithm processes a pseudo-random mixing function to deduce\nthe NextXmtTime of the current node. In fact if the pseudo-random mix of\nthe local node is superior to all the mixes of eligible competing nodes, the\nNextXmtTime for the local node is set to CandXmtOpportunityNum and the\nalgorithm returns a success. It is worth noticing that the proposed algorithm\nis fair and robust because all nodes are treated equally and scheduling seeds\nare varying pseudo-randomly for each frame leading to non-persistent\ncollisions.\nHowever, in uncoordinated distributed scheduling, the MSH-DSCH mes-\nsage is transmitted to the intended neighbor in the free slots of the data\nsub-frame without paying attention to possible collusions [10,11,28]. The\nMSH-DSCH message always includes the following fields [31]:\n■\nScheduling IE includes the next MSH-DSCH transmission time and\nhold-off exponent of the node and its neighbor nodes.\n■\nRequest IE conveys the resource request of the node.\n■\nAvailability IE implies the available channel resource of the node.\n■\nGrants IE conveys grant or confirm information of the channel\nresource.\nBoth centralized scheduling and distributed scheduling use the three-\nway handshake, which principle is given by Figure 2.17. If no MSH-DSCH\nis received for an uncoordinated distributed scheduling request, the second\nrequestee sends an MSH-DSCH:Grant packet.\n" }, { "page_number": 96, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n85\n1\n2\n3\n4\nRequester\nRequestee 1\nRequestee 2\nMSH−DSCH: Request\nMSH−DSCH: Grant\n(Confirmation and informs\nrequester’s neighbors) \nMSH−DSCH: Grant\nIf no MSH−DSCH: Grant packet\nis overheard for this uncoordinated\nrequest, then Requestee 2 sends\nMSH−DSCH: Grant packet here\nMSH−DSCH: Grant\nFigure 2.17\nThe three-way handshake.\nTransmission errors are corrected interactively, thanks to the Automatic\nRepeat Request (ARQ) protocol. The ARQ principle states that when a re-\nceiver detects corruptions in a message, it automatically requests a retrans-\nmission; then, after getting the correspondent ARQ message, the sender\nretransmits the message until it is correctly received or until the number of\nattempts exceeds a configured threshold. The ARQ mechanism is defined\nat the MAC layer; its implementation is optional and may be per-connection\nbased [47]. However, a connection cannot support ARQ and non-ARQ at\nonce.\n2.5.4.2\nHand-Over\nAn Access Service Network (ASN) includes at least one ASN gateway (GW)\nand a BS associated with one or more ASN gateway. The BS or ASN GW are\ncalled a serving BS or a serving ASN GW, respectively, when they manage\nthe MS before the hand-over and a target BS or a target ASN GW, respec-\ntively, if they are associated to the MS after the hand-over. Furthermore,\nan ASN GW can be an anchoring ASN GW when it used to relay MS data\nto the serving ASN GW. In this case, the CSN does not carry information\nabout the MS location and the IP address changes become less frequent.\nMobility management needs the implementation of hand-over proce-\ndures combined with the SS’s context management and data transmissions.\nFor instance, the data path function establishes the correspondent paths and\nguarantees the data transfers while the SS’s context and its exchange in the\nbackbone are handled by the context function. The hand-off functions are\nresponsible for the hand-over signaling and decisions. In fact, the hand-\nover procedure is first initiated by a request emitted by a serving hand-off\nfunction; then the involved targets reply and wait for the correspondent\nconfirmation. Only the entity which receives the confirmation becomes the\nserving one.\n" }, { "page_number": 97, "text": "86\n■\nSecurity in Wireless Mesh Networks\nIntra ASN hand-overs which take place between BSs belonging to the\nsame ASN do not result in important delays and data loss; moreover, they\ndo not induce changes in IP addresses because the movement of the SS is\ntransparent outside. However, inter-ASN hand-overs which occur between\nBSs belonging to different ASNs require a special coordination between\nthe involved ASN GWs where anchoring and re-anchoring are adopted. SSs\ncollect the channel information of the neighboring BSs either by performing\nranging or by listening to the current BS’s broadcast messages.\n2.5.4.3\nPhysical Layer Overview in WiMAX Mesh Mode\nThe IEEE 802.16a standard extends the physical layer defined for the 10 to\n66 GHz range to support mesh mode operations in the 2 to 11 GHz band\nof licensed and unlicensed spectrum [36]. In fact, the standard has enabled\nnon-line-of-sight (NLOS) operations while addressing the resulting multi-\npath constraint by adopting the OFDM modulation. Data bits enter the\nchannel coding block to be treated by the Forward Error Correction (FEC)\nand then interleaved [34]. They are then passed to the constellation map\nof the modulator. An Inverse Discrete Fourier Transform (IDFT) of length\nN is then applied to the data sequence, resulting in a frequency domain\nrepresentation bn composed of N carriers. A digital/analogical conversion\nis then applied and the resulting signal is low-pass filtered and modulated\nup to the carrier frequency of choice. The time domain impulse response\nof a multipath transmission channel approximates that of the Rayleigh\ndistribution [36].\nUsing the OFDM modulation allows a good average signal-to-noise ratio\n(SNR), but the SNR of each carrier varies widely. To address this issue,\nforward error correction codes are used. However, it is important to notice\nthat using OFDM in a noisy environment such as an NLOS air-link simplifies\nthe equalizer design and allows the demodulator estimating the SNR for\neach carrier and feeds this information to the FEC stage to squeeze the\nmost out of the channel [33].\nThe IEEE 802.16-2005, also known as IEEE 802.16e or Mobile WiMAX,\nwhich was approved in December 2005, is an improvement of the mod-\nulation schemes adopted by the original fixed WiMAX standard. In fact, it\nuses a new modulation method called Scalable OFDMA, which improves\nNLOS coverage by using advanced antenna diversity schemes and hybrid\nautomatic retransmission request. Moreover, the standard improves indoor\npenetration and introduces high-performance coding such as Turbo Coding\nto enhance security and NLOS performance.\n2.5.4.4\nQoS Support\n■\nQoS Support in WiMAX Mesh Mode: The IEEE 802.16 standard pro-\nvides QoS for the PMP mode by defining four classes of service:\n" }, { "page_number": 98, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n87\nunsolicited grant, real-time polling, non-real-time polling, and best\neffort. When examining the MAC header, we find a 16-bits field\ncalled CID, which is in charge of distinguishing between unicast\nand broadcast frames, defining service parameters, and identifying\nlink IDs. Figure 2.18 illustrates the CID of a unicast packet containing\nthe fields Reliability, Priority/Class, and Drop Precedence.\nThe Reliability field is set to zero when there is no retransmis-\nsion. It is set to one to indicate retransmit more than four times. The\nPriority/Class value indicates the priority of the packet and Drop\nPrecedence refers to the probability of the packet when conges-\ntion occurs. These three QoS parameters are defined in the protocol\ndespite the lack of a slot allocation algorithm that uses them. To\nachieve QoS features in the mesh mode, a simple slot allocation\nalgorithm has been proposed in [30]. The principle is to determine\na reasonable transmission time by looking up the channel resource\ntable after receiving a request and returning the detail of slot occu-\npation information. For this purpose, the node first computes the\nnumber of mini slots (R) requested for transmitting within a frame,\naccording to its Demand Level and Demand Persistence. Then, it\ndeduces the value of the next MSH-DSCH transmission time (T) by\nconsulting the neighbor table, which is stored locally. After that, the\nnode looks up R continuous available mini slots at the same posi-\ntion of the continuous frames (the number is Demand Persistence)\nstarting from time T. In case of success, it returns a grant to the\nrequester; otherwise, failure information is forwarded.\nUnfortunately, this simple algorithm is not sufficient for guaran-\nteeing the QoS. To improve it, the authors of [30] have set a check-\npoint along the first available time slots and a threshold in the chan-\nnel resource table. The number of allocated mini slots reflects the\nutilization of the data sub-frame in a certain degree and the thresh-\nold varies between 0 and 256. When the utilization level of the data\nsub-frame at checkpoint is lower than the threshold, the network\nstate is assumed good and the transmission requests will be treated\nwith the same priority. A utilization level higher than the threshold\nType\nReliability\nPriority\nclass\nDrop precedence\nXmt link ID\n16 bits\n8\n16 bits\n0\n2\n3\n6\nFigure 2.18\nThe CID field of a unicast packet.\n" }, { "page_number": 99, "text": "88\n■\nSecurity in Wireless Mesh Networks\nreflects a congested state. In this case, low-priority requests will be\nanswered by failure information.\nThe drawback of the improved algorithm is that one checkpoint is\nnot enough and may cause mistakes under some circumstances. To\naddress this issue, a second checkpoint is added. When the utiliza-\ntion level at checkpoint 1 is lower than the threshold, the algorithm\nturns to check the utilization level at checkpoint 2; if exceeded, it\nsearches a frame from checkpoint 2 whose utilization level is below\nthe threshold and allocates mini slots for the frame.\n■\nQoS Provision on the Backbone: Mesh routers forming the back-\nbone relay traffic between the client nodes and the wireline gate-\nways to communicate with external networks such as the Internet.\nTo increase the coverage area, new wireless routers may be eas-\nily added; however, an efficient QoS routing should be provided\nwhile addressing scalability issues and taking advantage of the low\nmobility and power consumption of the nodes. To address these\nissues, authors in [68] have presented a wireless DiffServ architec-\nture for the wireless mesh backbone. In fact, the DiffServ approach\nmay interconnect heterogeneous wireless/wireline networks; how-\never, its wireless version, which is proposed over the wireless mesh\nbackbone, needs to address the following challenges [68]:\n■\nRouters need to support both edge and core functionalities as\nthey may collect service requirements from different clients and\naggregate them to a unique service level agreement (SLA)\nrequirement or relay traffic to and from the gateways.\n■\nThe centralized bandwidth broker (BB), which collects traffic\nstatus at the edge/core router and monitors resource alloca-\ntion and QoS provision, cannot be defined in the mesh context;\ntherefore, a distributed protocol should be defined to guarantee\nthe BB services in a distributed manner.\n■\nThe wireless DiffServ should handle a large number of gate-\nways. Therefore, the service requirement from a wireless mesh\nbackbone represents the summation of all the aggregating SLAs\nthrough all the involved gateways. SLA configuration on each\ngateway should take into account the wireless mesh backbone\ntopology and the traffic density generated by each router.\n■\nWireless links capacity changes constantly. Therefore, the phys-\nical and link layers should be taken into account when perform-\ning QoS provisioning.\nMulti-hop networks generally adopt distributed control and\nresource allocation protocols. Therefore, the routing protocols\nare QoS-aware; they search for paths satisfying multiple QoS\nconstraints such as delay and bandwidth. The mesh backbone\nis a multi-hop network characterized by a low mobility scheme.\n" }, { "page_number": 100, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n89\nThe involved routers provide a broadband wireless connectivity\nand perform the differentiation and classification of the flows\ngenerated by their associated networks while optimizing the re-\nsources utilization. As a router may monitor multiple ad hoc\nnetworks or WLANs within its coverage area, it aggregates flows\ninto classes and routes, the flows of the same class in a sin-\ngle path satisfying that class QoS requirements. Authors in [68]\npropose a cross-layer routing protocol based on four compo-\nnents: the load classifier, the path selector, the call admission\ncontrol routine, and the route repair routine. The load classi-\nfier determines whether the traffic load of a certain class is low,\nmedium, or high, then triggers the path selector to select the\nless-congested gateway and select a suitable path to that gate-\nway based on the Greedy Perimeter Stateless Routing protocol\n[69]. Thereafter, the destination gateway triggers a call admis-\nsion control procedure which has MAC contention awareness.\nThe route repair routine is started when the route to the desti-\nnation gateway breaks or when it can no longer meet the QoS\nrequirements. In this case, the path selector should select a new\npath from the breaking point in order to minimize the overhead.\nThe wireless mesh backbone can adopt either a CSMA/CA\nor a reservation-based MAC [68]. The CSMA/CA approach is\nwidely deployed in the WLAN context; however, it suffers from\npoor throughput and unfairness problems when applied in a\nmulti-hop environment. The reservation-based MAC approach\nis gaining increasing interest as it guarantees contention-free\ntransmissions, thanks to reservations. Nevertheless the channel\nreservation is a challenging issue, as it needs to be monitored\nin a distributed manner [3]. To optimize the MAC resource uti-\nlization, resources which are not used by the high-priority traffic\nclass should be assigned to the low-priority traffic class. When\nreservation-based MAC is used, additional control mechanisms\nneed to be defined to exploit the resources originally reserved\nfor other classes. Controversially, the CSMA/CA MAC approach,\nwhich is completely distributed, may become suitable for the\nwireless DiffServ after addressing the hidden terminal problem,\nas stated in [68]. To serve the most prior traffic first, the black\nburst contention scheme is adopted to modify the traditional\nEnhanced Distributed Control Function (EDCF) proposed by the\nIEEE 802.11e standard. In fact, each node that wants to transmit\nshould first wait for the channel to be idle for an arbitration inter-\nframe period (AIFS) proper to its traffic class. Then, instead of\ntraditionally waiting for the back-off duration, the node should\nsend a black burst, the length of which (in the unit of slot time)\n" }, { "page_number": 101, "text": "90\n■\nSecurity in Wireless Mesh Networks\nequals the back-off timer in order to jam the channel. The node\nwill then wait for the channel to become idle. If it is the case,\nthe node may monitor the channel; otherwise, it will quit the\ncurrent contention, change the back-off duration, and wait for\nthe channel to be in an idle state for the AIFS again. The node\nwhich has high-priority traffic will have a long back-off timer\nso that the low-priority nodes will sense the black burst of the\nhigh-priority node and find the channel busy, thus being obliged\nto differ the transmissions.\n2.5.5\nDeployed Solutions\nConstructors such as Tropos Networks, Strix, and Nortel have already de-\nployed metropolitan mesh networks in the United States and Taiwan. This\nsection is an overview of the proposed coverage solutions.\n2.5.5.1\nTropos® Networks\nTropos Networks tries to offer data communications anywhere, anytime, to\nanyone that needs it. To achieve this goal, the Tropos MetroMesh architec-\nture combines the ubiquitous coverage of cellular with the ease and speed\nof WiFi. Thanks to this marriage, effects of interference and multi-path\nfading across the mesh are overcome while throughput in the range of > 1\nMbps (symmetric) is consistently delivered to standard WiFi client devices.\nMany cities in the United States have adopted the MetroMesh architec-\nture to deliver ubiquitous broadband access to their residents. The pioneer\ncase studies of Chaska and Corpus Christi deserve to be investigated.\n■\nThe Chaska Wireless Internet Service Provider: Chaska, Minnesota,\nhas always tried to offer attractive services to its residents. First, the\ncity started its own electricity utility so that its habitants have escaped\nthe pricing demands of a private utility. In 1998, the incumbent\ntelecommunications providers were ignoring the broadband data\nneeds of the schools in the community. To face the problem, the\ncity formed chaska.net, a WISP owned and operated by the city. The\nWISP implemented wireless point-to-multi point (PMP) technology\nto replace the traditional T-1 line required by the city’s educational\ninstitutions.\nBut the spring of 2004 was the real turning point in Chaska’s\nhistory. While more and more residents were asking for lower-priced\nbroadband and Internet connectivity that did not tie up phone lines,\nthe city government was struggling to attract new residents and\n" }, { "page_number": 102, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n91\nbusiness to Chaska, and to keep them in town rather than going\nto neighboring Minneapolis. After carefully considering the situa-\ntion, chaska.net decided to adopt the metro-scale WiFi from Tropos\nNetworks. The city’s wireless metropolitan network made use of the\ncity’s existing fiber network and was constructed using a combina-\ntion of Tropos Networks’ MetroMeshTM architecture, KarlNet PMP\nwireless backhaul connections, and an Operations Support System\n(OSS) from Pronto Networks. The deployment of wireless broad-\nband needed a capital investment of $535,000 and occurred in less\nthan eight weeks, although traditional wireline broadband networks\nand incumbent wireless (3G) networks can take years and require\ntremendous investments.\nAs it uses the 802.11 standard (WiFi) for backhaul and client\naccess, the network requires no proprietary radio frequency (RF)\nequipment for access devices. Besides, mobile users pay only $15.99\nper month with no time-term contracts required and have the abil-\nity to freely roam throughout the entire 16 square miles of the city\nbecause the 230 deployed Tropos 5110 MetroMesh routers allow\ntransparent roaming. Backhaul was injected at 36 locations around\nthe city using a combination of KarlNet PMP wireless links and\nconnections to the city’s fiber network. Scalability was guaranteed\nbecause the Tropos 5110 MetroMesh routers automatically reorga-\nnize to take advantage of the increased capacity and the additional\nbackhaul.\nBy using Tropos Networks’ metro-scale WiFi technology and exist-\ning infrastructure, chaska.net provides broadband access to all 7500\nhomes in the city as well as city employees, public safety officials,\nand small businesses at rates up to 60 percent less than competing\nbroadband services, and in many cases at or below the cost of dial-\nup services. The subscriber management is done using the Tropos\nControl element manager, which allows chaska.net staff to monitor\nthe WiFi network from a centralized location. When subscribers ac-\ncess the network, the Pronto OSS redirects them to a Web page on\nthe chaska.net Web server. In fact, the Pronto OSS platform and Com-\nmunity Broadband Gateway are in charge of provisioning, authenti-\ncation, customer billing, administration, customer relationship man-\nagement (CRM), and roaming agreements. In addition, a global MAC\naddress\nwhite\nlist\nis\ndefined\nto\nprovide\nadditional\nsecurity\nsupport.\n■\nThe Multi-Use Metro-Scale WiFi, City of Corpus Christi, Texas:\nCorpus Christi is rated as the largest city on the Texas coast and\nthe nation’s sixth largest port. The city always relied on its technol-\nogy infrastructure to enhance the productivity and efficiency of its\n" }, { "page_number": 103, "text": "92\n■\nSecurity in Wireless Mesh Networks\nmunicipal services, attract more business, and better serve its res-\nidents. However, Corpus Christi was facing permanent problems\nwith meter reading. “Meter readers often have difficulty accessing a\nproperty because of fences or dogs,” explained Leonard Scott, MIS\nunit manager and program manager for the WiFi project. “We av-\nerage several complaints per day, every day, from customers who\nbelieve their utility statements are incorrect. If someone wants to\nbuy a house, there is no easy way to check gas and water usage\nhistory.” To address this issue, Corpus Christi has decided to auto-\nmate meter reading for municipal gas and water services that supply\na 147-square-mile area.\nAlthough a fiber-optic network backbone was covering two-thirds\nof the city, it did not extend to the third of the area that the Auto-\nmated Meter Reading (AMR) system would need to cover. To allow\ncoverage of the totality of the zone, Corpus Christi selected Tropos\nNetworks for relaying gas and water meter data from AMR concen-\ntrators to the city’s utilities business office system. With automated\ndata collection, gas and water customers were able to check daily\nmeter data online and view a property’s gas and water consumption\nhistory while the municipality was better able to monitor gas usage\nand water flow.\nAfter living the success story of the AMR application which used\na limited portion of the available bandwidth, the city departments\nsoon predicted the potential for hosting new services such as vehi-\ncles equipped with laptops for police, fire, and other public safety\nofficers; mobile desktops for field supervisors and managers; and\nanywhere, anytime access for residents and visitors to city resources\nsuch as the library, City Hall, and museums. The only critical ques-\ntion was how to allow broad use of the wireless network while\nrestricting the municipal system to some authenticated users and\nguarantee the security services for the public safety system. To over-\ncome this problem, the mesh metro-mesh architecture powered by\nTropos Networks was combined with the Pronto’s OSS, which pro-\nvides an SSL-encrypted registration and authentication process and\nsupports VPN, which allows secure and encrypted access. Besides,\nthe 300 Tropos 5110 outdoor MetroMesh routers allowed the deliv-\nery of multimedia data with automated roaming over the coverage\narea.\nThanks to the combination of the metro-mesh architecture pow-\nered by Tropos Networks with the OSS for subscriber management,\nCorpus Christi’s residents, municipality officers, public safety agents,\npublic works department employees, and building inspectors have\nbeen able to get broadband ubiquitous access to vital online infor-\nmation while they are in the field [27].\n" }, { "page_number": 104, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n93\n2.5.5.2\nStrix Systems\nThe Access/One Network Outdoor Wireless System (OWS) of Strix Systems\nis designed for the deployment of 802.11 networks across large urban ar-\neas, rural counties, and entire regions. OWS solutions have been deployed\nin hundreds of networks worldwide, outdoor and indoor, for the metro,\npublic safety, government, energy, transportation, hospitality, education,\nenterprise, residential, and carrier access markets. The resulting structured\nwireless mesh networks provide intelligence, scalability, security, and un-\nrivaled performance. Using Access/One, public safety markets can deploy\nsecure and manageable wireless networks in unlicensed spectrums that\nsupport voice, video, and data applications. Furthermore, high-speed Inter-\nnet access can be provided even in underserved rural areas.\n■\nThe Tempe Case Study: The City of Tempe, Arizona, selected the\nAccess/One Network OWS for its citywide WiFi deployment [32].\nTempe will offer secure WiFi access for its residents, businesses, and\nvisitors. Moreover, public safety agents will be provided with WiFi\naccess to their secure private network within all 40 square miles of\nthe city limits. Strix was chosen in partnership with MobilePro for the\nhigh throughput and low latency the system offers across large net-\nworks. When complete, the citywide network will provide anytime,\nanywhere access to residents, businesses, and municipal workers,\nenhancing the way people connect to the Internet, do business, and\nserve the community.\nThe City of Tempe was considered validation of Strix’s technol-\nogy because it was hand-selected from a group of 113 possible pro-\nposals. This also speaks very highly of the combined systems and\nservices that the solution is capable of deploying. Some experts af-\nfirm that the Access/One Network OWS is an efficient solution that\nenables customers to dedicate radios for both ingress and egress in\nthe mesh backhaul as well as separate radios for client access.\n■\nThe Chittagong Case Study: Strix Wireless Mesh will enable new-\ngeneration wireless voice/data/video services for 3.5 million people\nin Chittagong, the commercial capital of Bangladesh. The deployed\nmesh network will be based on Strix’s Access/One Network OWS\nand will provide broadband phone and Internet service for residents,\nbusinesses, and visitors. Accatel Inc. has partnered with Nextel Tele-\ncom to deploy the citywide wireless mesh infrastructure; it is now\ninstalling 90 Strix OWS nodes for the initial network deployment,\nwhich will support 10,000 voice subscriber lines in an eight-square-\nmile area. The second phase of the project will add 15 to 20,000\nvoice subscribers in 12 months. Within three years, the Strix wireless\nmesh network is expected to include hundreds of OWS nodes and\nserve 200,000 voice subscriber lines. In the near future, the wireless\n" }, { "page_number": 105, "text": "94\n■\nSecurity in Wireless Mesh Networks\nmesh network will be deployed over the whole area of Chittagong\nand other cities within the licensing area.\n■\nNortel’s Case Study: Marshalltown, Iowa, is a rural community with\na small population. To encourage economic development and at-\ntract businesses and residents, Marshalltown has decided to adopt\nthe wireless technology and launch the first WiFi city network in\nthe state of Iowa. The Marshalltown Economic Development Impact\nCommittee, in conjunction with critical communications system\nintegrator RACOM, has chosen Nortel’s wireless mesh solution to\ninitially provide end-user WiFi services to a 20-square-block area in\nthe downtown core. The network infrastructure is based on seven\nNortel 7220 WLAN APs supported by a Nortel Wireless Gateway\n7250, giving free public WiFi services to local residents and busi-\nnesses. The new broadband network delivers mobile Internet access\nat 800 kbps for roaming users within the downtown core. Public\nsafety workers are also supported by the network. Besides, the mesh\nsolution allows the network to differentiate high-priority emergency\nresponse traffic from low-priority public Internet access. Marshall-\ntown plans to support the delivery of data communications for emer-\ngency response teams, including video surveillance, as well as ac-\ncess to local, state, and national databases for relevant information.\nIn the near future, the wireless mesh network will cover the entire\ncounty and support WLAN IP telephony and VPN capabilities [39].\n2.6\nWireless Mesh WAN\nMesh WANs intend to provide ubiquitous mobile broadband wireless\naccess in a cellular architecture while supporting mesh networking in\nindoor and outdoor scenarios. For instance, mobile travelers can enjoy\nInternet access while passengers information services, remote monitoring\nof in-vehicle security video, and driver communications may be supported\nwithin a complete transportation system. Besides, the guarantee of an NLOS\ncommunication enables users to extend the coverage area and to build\na wide mesh network that provides Internet-based applications such as\nstreaming and VoIP with enhanced throughput, reliable services, and QoS\nsupport.\nThe Mobile Wireless Broadband Access (MWBA) is a transmission tech-\nnology that allows important throughput for last-mile wireless connections\n[43], which is why it has been adopted by both IEEE 802.20 and IEEE\n802.16e standards. Broadband services are provided to potential customers\nwith support of multimedia applications. Besides, MBWA systems are resis-\ntant to rapid channel variation and address the implications of mobility on\nthe IP layer by maintaining the routability of packets during IP hand-off.\n" }, { "page_number": 106, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n95\nThe IEEE 802.20 standard intends to provide wireless access systems\nwith mesh networking support for high-speed mobile subscriber stations\nwithin a medium-to-extended metropolitan area. In fact, IEEE 802.20 oper-\nates in licensed bands below 3.5 GHz and specifies the MAC and physical\nlayers extensions that offer ubiquitous mobile broadband access for cellular\nand mesh architectures for mobile users traveling at up to 155 mph with\nNLOS communications support. In the following sub-sections, what little\ninformation is currently available about the 802.20 PHY and MAC layers\nwill be presented and the similarity and differences with respect to 802.16e\nwill be discussed.\n2.6.1\nIEEE 802.16 Mobility Management\nThe IEEE 802.16e standard is an amendment of the IEEE 802.16d standard,\nalso known as IEEE 802.16-2004, which supports the mesh mode. IEEE\n802.16e adapts the scalable OFDMA (SOFDMA) technique at the physical\nlayer to improve multi-access capabilities while enhancing the MAC layer by\naddressing mobility issues and particularly hand-over. IEEE 802.16e over-\nlaps with the mandate of IEEE 802.20 and introduces nomadic capabilities\nallowing mobile users to connect to wireless Internet services providers\nwhile moving at a speed of 75 to 93 mph. To manage client mobility, dif-\nferent types of hand-over have been addressed [48]. Following is a brief\ndescription of each type.\n■\nMS-Initiated Hand-Over: This hand-over occurs when a node detects\ndegradations in the signal with its serving BS or when it deduces\nthat it can get a higher QoS at another BS. The hand-over deci-\nsion is taken after collecting gain information from the neighboring\nnodes which periodically broadcast the mobile neighbor advertise-\nment message specifying frequency of the BS they belong to, its\nidentifier, the types of services it supports, and its available radio re-\nsources. The mobile station may also precede a neighbor scanning\nby synchronizing with some targeted BS’s downlink transmissions\nand estimating the quality of the physical channel. After defining a\nlist of candidate BSs, the MS sends a notification to its serving BS. The\nserving BS coordinates with the candidates to get a hand-over pre-\nnotification response and define a list of targets. The MS may then\nchoose one target and should inform its serving BS that it is leaving.\n■\nBS-Initiated Hand-Over: A serving BS may decide to exclude some\nMSs when it detects that the managed nodes are leaving the cov-\nerage zone or when it estimates that it can no longer provide the\nrequired QoS.\n■\nSoft Hand-Over: Soft hand-over is performed when an MS is able to\nreceive the same MAC/PHY protocol data units from one or more\n" }, { "page_number": 107, "text": "96\n■\nSecurity in Wireless Mesh Networks\nBSs, thanks to diversity combining at the antenna. Soft hand-over\npermits the MS to continue receiving real-time data despite the hand-\nover procedure; however, it requires multiple antennas and it is more\ncomplex.\n2.6.2\nIEEE 802.20\nThe IEEE 802.20 standard intends to provide a downlink rate of 1 Mbps and\nan uplink one of 300 kbps for high-speed mobile users while guarantee-\ning efficient packet-based data services with real-time traffic support [44].\nIt supports the mesh networking paradigm and the NLOS communications.\nThe architecture of an IEEE 802.20 network guarantees seamless integration\nof different user domains. In fact, targeted applications are VoIP, financial\ntransactions, online gaming, audio and video streaming, videoconference,\nWAP, file download, Web browsing, etc. The supported devices (laptops,\nPDAs, and smart phones), which have different mobility, battery, and stor-\nage constraints, will generate different traffic and application models, de-\npending on their characteristics. However, they will benefit from a seamless\nubiquitous access.\nThe IEEE 802.20 standard gives the specifications of the physical and\nMAC layers that provide enhanced services to the third layer of the OSI\nmodel to achieve reliable IP packets routing between external terminals\nand mobile users or between mobile users. The IEEE 802.20 MWBA system\narchitecture addresses resource allocation, rate management, and authen-\ntication issues, and pays specific attention to location management and\nhand-over.\nTable 2.1 summarizes the principal characteristics of the air interface as\nspecified by the IEEE 802.20 standard. In addition to its support for the\nmultimedia applications and QoS requirements, IEEE 802.20 guarantees a\nseamless hand-over between other network technologies, thanks to the\nadaptation layer (virtual interface). In fact, the hand-off is implemented at\nthe MAC layer while the virtual interface manages multiple wireless network\ninterfaces on a single host by providing a virtual MAC address to the station.\nAs a result, each mobile node is assigned a unique IP address although it\nmay move between different wireless networks; the station’s mobility will\nbe reflected by the changes in the virtual MAC values.\n2.6.2.1\n802.20 PHY Layer\nThe PHY layer of the 802.20 standard is typically based on the technolo-\ngies developed in the 802.16 working groups. The standard for the PHY\nlayer, however, is more heavily angled toward use in a mobile setting and\nseems to be inclining toward using OFDMA (Orthogonal Frequency Divi-\nsion Multiple Access) in a similar way to 802.16e. This mainly can reduce the\n" }, { "page_number": 108, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n97\nTable 2.1\nThe IEEE 802.20 Air Interface Specifications\nCharacteristic\nTarget Value\nMobility\nVehicular mobility classes up to 250 kmph\n(as defined in ITU-RM.1034-1)\nPeak user data rate (downlink [DL])\n> 1 Mbps\nPeak user data rate (uplink [UL])\n> 300 kbps\nPeak aggregate data rate per cell (DL)\n> 4 Mbps\nPeak aggregate data rate per cell (UL)\n> 800 kbps\nAirlink MAC frame RTT\n< 10 ms\nBandwidth\ne.g., 1.25 MHz, 5 MHz\nCell sizes\nAppropriate for ubiquitious MANs and\ncapable of reusing existing\ninfrastructure\nMaximum operating frequency\n< 3.5 GHz\nSpectrum (frequency arrangements)\nSupports FDD and TDD frequency\narrangements\nSpectrum allocations\nLicensed spectrum allocated\nto the mobile service\nSecurity\nSupport AES\ndevelopment time of products. However, the possibility of using OFDMA\n(Orthogonal Frequency Division Multiple Access) on the downlink connec-\ntion and CDMA (Code Division Multiple Access) on the uplink has been\nmentioned. The reason for using CDMA on the uplink is that using OFMDA\nsomewhat limits the benefits that antenna technologies like spatial multi-\nplexing can provide. CDMA can help to reduce this limitation by assigning\nthe same bandwidth resources to all users in a sector and using spatial\nprocessing at base station to recover the signal [26].\nModulation and coding in 802.20 is essentially identical to that of\n802.16a/d. Besides, to allow flexible high-speed mobility, the 802.20 stan-\ndard is expected to support basically all of the advanced transmission op-\ntions that the 802.16 standards define. These standards include, but are\nnot limited to space–time block code and various forms of spatial multi-\nplexing/MIMO (Multiple-Input Multiple-Output). A wide variety of channel\nbandwidths from 1.25 to 40 MHz are also expected to be supported with\nboth TDD and FDD multiplexing. Using 1.25 MHz channel speeds (similar\nto ADSL), while providing 1 Mbps downstream and 300 kbps upstream, are\nexpected to scale with wider channels. This will allow the support of up\nto 100 users per cell.\n2.6.2.2\n802.20 MAC Layer\nThe MAC layer of the 802.20 standard is also loosely based on technolo-\ngies developed in the 802.16 working groups. Similar to 802.16, the 802.20\n" }, { "page_number": 109, "text": "98\n■\nSecurity in Wireless Mesh Networks\n802.3\nPhysical\n802.4\nPhysical\n802.11\nPhysical\n802.12\nPhysical\n802.3\nMAC\n802.11\nMAC\n802.12\nMAC\n802.4\nMAC\nPhysical\nlayer\nData\nlink\nlayer\n802.1 Bridging\n802.2 Logical link control\n802.1 Management\n802 Overview &\narchitecture \n802.10 Security\nFigure 2.19\nLLC functionalities.\nMAC is divided into convergence-specific and common-part sub-layers. Fur-\nthermore, mobility techniques developed in 802.16e such as hand-off and\npower management are also implemented in the 802.20 standard. Figure\n2.19 details the logical link control (LLC) services that intend to guaran-\ntee reliable data transmissions. It also shows that the IEEE 802.20 may\nsupport common and specific parts of the physical layer to support vari-\nous PHY technologies [45]. The connection establishment mechanism to be\nprovided by 802.20 is not yet fully defined, but due to the standard’s resem-\nblance to 802.16e, it is expected that the mechanisms will be largely similar.\nOne difference between the two, however, is that CDMA (with respect to\nOFDM/OFDMA) could be utilized on uplink connections.\nBecause 802.20 is a fully mobile standard, it will provide support for all\ntypes of hand-off mechanisms to enable users to freely roam between cells\nwithout interruption. Soft hand-off provision will be entirely integrated.\n802.20 also will fully integrate higher-level hand-offs over Mobile IPv4 and\nMobile IPv6. Because different forward and reverse-link connection mech-\nanisms may be used, hand-off will need to occur in both directions [26].\nThe level of QoS support that 802.20 will offer is to some extent unde-\ncided at this moment. The common requirements document agrees, how-\never, on the fact that DiffServ and RSVP will be supported for end-to-end\ncompatibility with traditional networks. Finally, note that the 802.20 stan-\ndard offers performance similar to that provided by usual 2.5G and 3G\ncellular technologies. 802.20, however, presents the clear advantage of be-\ning a fully IP-based, packetized network standard. Consequently, network\nthroughput is enhanced versus a circuit-switched standard, because mes-\nsages do not have to be encoded from pre-allocated circuits into packets\n(and back) each time a request is sent or received. Additionally, the 802.20\noffers a higher spectral efficiency than any current cellular standard. Thus,\nthe 802.20 is expected do more with less channel bandwidth and would\nhandle a higher number of users per cell.\n" }, { "page_number": 110, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n99\n2.7\nAdvanced Issues\nFactors such as network topology and architecture, traffic nature, and node\nmobility highly mark the mesh network’s capacity and performance, thus\naffecting protocols development and implementation. All protocols need\nto be improved or reinvented while considering a cross-layer design. This\nsection gives an overview of the hottest research issues aimed at designing\nscalable, low-cost, and easily deployable wireless mesh networks.\n2.7.1\nPhysical Layer\nWMNs physical layer should be revised to provide important rates and wide\ncoverage while enhancing reliability by solving the fading, multipath, and\ninterference constraints. Traditional modulation techniques such as OFDM\nand UWB should be replaced by new schemes that allow better data rates\nin larger areas. For instance, the MIMO technique, which intends to im-\nprove the wireless network capacity by adopting antenna diversity and\nspatial multiplexing, can be exploited. In fact, using multiple antennas for\nreception provides the receiver with replicas of the transmitted signal, thus\nreducing fading and interferences. Moreover, adopting spatial multiplexing\npermits the simultaneous transmission of different data streams by breaking\nthe channel into multiple spatial channels and then using each of them to\ntransmit a differently encoded traffic.\nAs diversity techniques are inefficient in case of strong interference,\nsmart antennas with beam-forming capability may also be adopted to pro-\nvide the receiver with high gain in the direction of the desired signal and\nlow gain in all other directions. Cheap directional-antenna implementation\nand frequency-agile techniques should be further investigated to build a\nhigh-capacity wireless backhaul system [62]. The MAC layer design should\nalso be done according to the added values of the physical layer to achieve\nthe expected improvements. Many MAC protocols as stated in [63–66] have\nbeen developed to support directional and smart antennas in the ad hoc\nnetwork context, but an additional effort is required to implement a MAC\nprotocol with multi-antenna-systems support. Moreover, cognitive radios\ntechnologies represent a new research field that needs to be investigated.\n2.7.2\nMAC Layer\nMesh nodes mobility and nature (router or client) combined with power\nconstraints add complexity to the design of a MAC scalable protocol. In\nfact, existing medium access-control protocols (such as CSMA/CA), which\napply to the ad hoc context, suffer from poor performance and frequent\ncollisions when the number of nodes increases; therefore, they should be\nreplaced by TDMA and CDMA schemes while overcoming the induced\n" }, { "page_number": 111, "text": "100\n■\nSecurity in Wireless Mesh Networks\ndifficulties. Advanced techniques such as MIMO and cognitive radios, which\ncan be implemented at the physical layer, need a particular MAC design to\neffectively enhance the throughput and coverage.\nThe scheduling is also a critical issue because it should address multi-\nuser diversity according to the cross-layer design. In fact, transmission op-\nportunities allocation should be coordinated among all wireless routers\nto grant transmission to users experiencing peak in their channel quality\n[67]. Moreover, open research issues related to scheduling should deter-\nmine how to profit from other diversity techniques implemented at the\nphysical layer, such as spatial diversity and frequency diversity, to en-\nhance the throughput. Besides, interoperability of various wireless tech-\nnologies requires the definition of particular bridging functions at the MAC\nlevel. Furthermore, a multi-channel multi-transceiver MAC can be a promis-\ning solution to guarantee reliability and enhance the provided data rates.\nFinally, a better QoS has to be offered at the MAC level to support multi-\nmedia traffic transmissions that are particularly affected by delays, packet\nloss, and jitter.\n2.7.3\nNetwork Layer\nMulti-hop communication protocols rapidly lose their performance when\nthe network size increases. Routing schemes designed for WMNs should\nensure scalability and enhance network performance without adding com-\nplexity and management difficulties. In fact, the destination of mesh traffic\nmay be multiple mobile nodes; furthermore, the same traffic may simultane-\nously follow multiple paths to reach the same AP. Thus, the routing proto-\ncols need to rely on correct link status information provided by the physical\nand MAC layers to discover high-quality routes. New routing metrics that\nreflect the loss rate and the available bandwidth of intermediate links need\nto be developed. Multicast traffic routing can also be a hot research topic.\nCross-layer design, which intends to enhance routing performances by con-\nsidering MAC parameters and feedback, is a promising research issue that\nneeds to be further investigated. Routing protocols should also take into\nconsideration the mesh nodes’ nature (which can be routers or clients) to\ncorrectly respond to different mobility and power constraints.\n2.7.4\nTransport Layer\nTransport protocols that are used in the ad hoc context are also adopted\nby the WMNs. These protocols can be classified as reliable TCP variants,\nentirely new reliable protocols, or protocols designed for real-time delivery.\nTCP variant protocols aim at overcoming the performance degradations\nexperienced by TCP when it is applied to the ad hoc context. In fact,\nnon-congestion packet losses caused by the transmission over unreliable\n" }, { "page_number": 112, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n101\nwireless links are considered by TCP as congestion losses and induce severe\nthroughput decreases. To address this issue, the protocol designed in [50]\nadopts a feedback mechanism that allows a differentiation between losses\ncaused by congestion and those caused by wireless channels; however, a\nfuture study is needed to correctly design a loss differentiation approach\nand to accordingly modify the TCP protocol for WMNs.\nBesides, the connection-oriented TCP protocol which relies on ACK\nreception is highly affected by mesh network asymmetry in terms of band-\nwidth, loss rates, and latency [37]. In fact, TCP data and the correspondent\nACK may take different routes in the mesh network, thus leading to perfor-\nmance degradations. Some ACK processing schemes have been proposed\nand a different network architecture has been presented in [15] to solve\nthe asymmetry-related problem, but their effectiveness for WMNs should\nbe further investigated. A cross-layer optimization can also be adopted to\nenhance the TCP performance because the network asymmetry is closely\nrelated to lower-layer protocols. Moreover, the high variation of the RTT\ncaused by node mobility and dynamic path changes has severe conse-\nquences on the TCP performance. Adapting TCP to RTT variation in the\nWMNs is still an open research topic.\nTo address TCP shortcomings, new protocols have been developed. To\nthis end, the ATP protocol, [12], which is rate-based, differentiates between\ncongestion and non-congestion losses by examining the resulting delays\nand does not set transmission time-outs while addressing congestion con-\ntrol and reliability separately. However, adopting a brand new transport\nprotocol for the WMNs will result in non-interoperability with existing tech-\nnologies. More specifically, WMNs should be able to permit network access\nfor conventional and mesh clients and wireless mesh nodes which need to\naccess the Internet and also to be integrated with heterogeneous wireless\nnetworks such as IEEE 802.11, 802.16, and 802.15. One solution will be\nthe development of a special adaptive TCP variant for WMNs which ad-\ndresses traditional TCP performance degradations while being compatible\nwith the traditional TCP protocol. Furthermore, end-to-end real-time trans-\nmission guarantees have been addressed by both RTP (Real Time Protocol)\nand RTCP (Real-Time Transport Protocol) in compliance with an RCP (Rate\nControl Protocol). However, there has been no RCP proposition specifically\ndesigned for the WMNs.\n2.7.5\nApplication Layer\nNew application layer algorithms need to be developed so that real-time\nInternet applications can be supported by multi-hop wireless mesh net-\nworks. Furthermore, distributed information sharing over WMNs has\nspecific characteristics that need to be addressed by new applications\nprotocols. Finally, new applications that take advantage of the WMN’s\n" }, { "page_number": 113, "text": "102\n■\nSecurity in Wireless Mesh Networks\nparticularities need to be invented to effectively provide an added value.\nFor example, new tools may be developed for a home networking envi-\nronment to achieve home automation by allowing the remote monitoring,\nconfiguration, and control of all electronic devices.\n2.7.6\nNetwork Management\nA centralized control of node location and hand-over is not applicable in the\nmesh context where an LOS with the BS is not required and where the client\nnodes may constantly roam while mesh routers have restricted mobility.\nDeveloping a distributed location management scheme for WMNs is an\ninteresting research topic that needs to be investigated. In the same way,\npower management procedures vary according to the nature of the mesh\nnodes. On one hand, mesh routers which do not have power constraints\nneed to manage their transmission power to control the connectivity and\nreduce interference while increasing the spectrum spatial-reuse efficiency.\nOn the other hand, mesh clients which may be IP phones or sensors require\nparticular power efficiency.\nConsequently, power management for the WMNs is an open research\ntopic that needs to be further investigated. Finally, network monitoring\nprotocols need to be developed to effectively manage mesh routers and\nenhance network performance. In fact, mesh routers have to report statisti-\ncal data to one or more servers to detect network anomalies and correctly\nrespond to them. Special data processing algorithms need to be developed\nand network management procedures designed for the ad hoc networks\nneed to be further enhanced to support large-scale mesh networks.\n2.7.7\nSecurity\nSecurity schemes designed for WLANs provide authentication, authoriza-\ntion, and accounting services by implementing them at the AP or at special\ngateways. Besides, VPN techniques are provided over WLANs using stan-\ndard key encryption algorithms for tunneling, such as IPSEC. Unfortunately,\nsuch schemes are not completely suitable for WMNs because the WMNs\ndo not provide a trusted centralized party that ensures a secure key and\ncertificates management. Besides, attackers may easily benefit from the lack\nof infrastructure to target routing and MAC protocols, leading to congestion\nand denial of service.\nAll these security breaches need to be addressed to convince wireless\nmesh networks customers to subscribe to reliable services. Security mech-\nanisms need to be embedded into the communications protocols of the\ndifferent layers so that intrusions are detected and tolerated. Designing\n" }, { "page_number": 114, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n103\na cross-layer framework that monitors the security of the communication\nprotocols is a challenging research topic that needs to be investigated.\n2.8\nConclusion\nThe goal of this chapter has been to present the wireless mesh networking\nfundamentals aimed at designing scalable, low-cost, and easily deployable\nmesh networks with coverage ranges from PAN to WAN. We may state that,\nalthough they inherit from the MANETs characteristics, mesh networks have\ntheir own specificities. In fact, scalability issues need to be addressed as the\nnetwork may integrate a large number of nodes and provide a wide cov-\nerage. Besides, distributed protocols need to be implemented to guarantee\nan efficient network management and control. As multimedia applications\nsupport is a must, mesh networks need to rely on QoS-aware routing pro-\ntocols able to establish the most suitable path while providing the QoS\nrequirements in terms of bandwidth, delay, and jitter. Nodes mobility man-\nagement and hand-off should also be addressed because clients need to\nmove at different speeds without losing access to the applications they\nare using (e.g., Internet access, access to a public-safety private network,\netc.). Last but not least, mesh networks need to provide advanced security\nmechanisms to encourage client subscribing to reliable services.\nWe can state that mesh PANs, LANs, MANs, and WANs share common\ncharacteristics and face common communication challenges although their\nrequirements may differ. For instance, when addressing transmission issues,\nwe can conclude that the UWB technique enhances the meshing capabil-\nities, but is only applicable in the short-range communications context.\nTherefore, different transmission techniques may be used in the MAN and\nWAN context to support node mobility at medium and high speeds while\nresisting multipath and fading. To provide QoS, it is possible to adopt\nthe IntServ approach for PANs and LANs because the node number is not\nvery important. However, a DiffServ approach fits the MANs and WANs\ncontexts because it provides a scalable solution and guarantees soft QoS\nrequirements. In addition, mobility constraints highly differ according to\nthe network size. In fact, in the mesh PAN context, it is difficult to maintain\nQoS-aware paths because both routers and mesh nodes are mobile; how-\never, the average speed is about 5 kmph. Mesh LANs always rely on a fixed\ninfrastructure; nevertheless, they need to address hand-over and roaming\nissues as the served mobile nodes may move from one ESS to another.\nMesh MANs and WANs include a fixed backhaul and a large number of\nmobile nodes moving at medium or high speed; therefore, guaranteeing\nQoS and addressing hand-over and roaming becomes a challenging issue,\nespecially when propagation conditions induce multipath and fading.\n" }, { "page_number": 115, "text": "104\n■\nSecurity in Wireless Mesh Networks\nMany works are currently being conducted on designing robust mesh\nnetworks ranging from PAN to WAN, but the finalized standards versions\nhave not yet been released.\nReferences\n[1]\nRaffaele Bruno, Marco Conti, and Enrico Gregori, Mesh networks: Com-\nmodity multihop ad hoc networks, IEEE Communications Magazine,\nMarch 2005, 43(3): 123–131.\n[2]\nMyung J. Lee, Jianliang Zheng, Young-Bae Ko, and Deepesh Man Shrestha,\nEmerging standards for wireless mesh technology, IEEE Wireless Commu-\nnications, April 2006, Volume 13, Issue 2, pp. 56–63.\n[3]\nIan F. Akyildiz, Xudong Wang, and Weilin Wang, Wireless mesh networks:\nA survey, Computer Networks, Elsevier, 2005.\n[4]\nDavid H. Axner, The Up Side and Down Side of Wireless Mesh Networks,\navailable at www.packethop.com/pdf/business communications review\njanuary 2006.pdf.\n[5]\nSteven Conner and Roxanne Gryder, Building a Wireless World with\nMesh Networking Technology, Technology@Intel Magazine, November\n2003, available at www.intel.com/technology/magazine/communications/\nnc11032.pdf.\n[6]\nBogdan Carbunar, Ioannis Ioannis, Cristina Nita-Rotaru, and Aaron Walters,\nBuilding Trustworthy Wireless Mesh Networks, Dependable and Secure\nDistributed Systems Labs, Purdue University (Microsoft presentation).\n[7]\nFrancis daCosta, Managing the Performance of Ad hoc Mesh Networks,\nMeshDynamics proposal for the IEEE P802.15 Working Group for Wire-\nless Personal Area Networks (WPANs), IEEE P802. 15-04-0211-00-0005,\nMay 2004, available at www.meshdynamics.com/Publications/MDWMAN-\nOVERVIEW.pdf.\n[8]\nJianliang Zheng, Yong Liu, Chunhui Zhu, Marcus Wong, and Myung Lee,\nIEEE 802.15.5 WPAN Mesh Networks, IEEE P802.15-05-0260-00-0005. Avail-\nable at grouper.ieee.org/groups/802/15/pub/ 05/15-05-0260-00-0005-802-\n15-5-mesh-networks.pdf.\n[9]\nJohn Boot, IEEE P802.15.TG5 Call for Applications, IEEE P802.15-05/267r1,\nMay 2004.\n[10]\nJack Lang, Mesh Network Outline, IEEE P802.15-15-03-0393-00-0030,\nSeptember 2003, available at grouper.ieee.org/groups/802/15/pub/03/15-\n03-0393-00-0030-mesh-network-outline.doc.\n[11]\nJinyun\nZhang,\nMesh\nNetworking\nSupport\nfor\nIEEE\n802.15\nWPAN,\nMitsubishi Electric Research Laboratories, 2005, available at http://www.\nmerl.com/projects/wpan-mesh/.\n[12]\nK. Sundaresan, V. Anantharaman, H.-Y. Hsieh, and R. Sivakumar, ATP: A\nReliable Transport Protocol for Ad hoc Networks, in ACM International\nSymposium on Mobile Ad hoc Networking and Computing (MOBIHOC),\n2003, pp. 6475.\n" }, { "page_number": 116, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n105\n[13]\nIEEE\nP802.15.4/D18,\nDraft\nStandard:\nLow\nRate\nWireless\nPersonal\nArea\nNetworks,\nFebruary\n2003,\navailable\nat\nwww.ember.com/pdf/\nEM2420datasheet.pdf.\n[14]\nAccess One Network, Strix Systems, Product description, available at www.\nstrixsystems.com.\n[15]\n3210 Indoor MetroMesh Router, Tropos Networks, available at www.tropos.\ncom.\n[16]\nDaniela Maniezzo, Gianluca Villa, and Mario Gerla, A Smart MAC-Routing\nProtocol for WLAN Mesh Networks, UCLA Technical report #040032,\nOctober 2004, available at www.cs.ucla.edu/ST/docs/tech rep2004.pdf.\n[17]\nWireless LAN Infrastructure Mesh Networks: Capabilities and Benefits,\nA Farpoint Group White Paper, Document FPG 2004-185.1, July 2004,\navailable at http://wireless.itworld.com/4260/farpoint wlanmesh/.\n[18]\nJoseph F. Herzig Jr., An Analysis of the Feasibility of Implementing Ultra\nWide-band and Mesh Network Technology in Support of Military Opera-\ntions, Thesis, Naval Postgraduate School, Monterey, California, March 2005,\navailable at http://www.stormingmedia.us/24/2422/A242234.html.\n[19]\nAre Two Links Better than One? Digitimes interviews Ted Kuo of Accton\nTechnology about the state of mesh standards.\n[20]\nRoxanne Gryder and Steven Conner, Building a Wireless World with\nMesh Networking Technology, available at www.intel.com/technology/\nmagazine/communications/nc11032.pdf.\n[21]\nQi Xue and Aura Ganz, QoS routing for mesh-based wireless LANs,\nInternational Journal of Wireless Information Networks, Vol. 9, No. 3, July\n2002.\n[22]\nMetro-Scale WiFi as City Service, chaska.net, Chaska, Minnesota, A Tropos\nNetworks Case Study, October 2004, available at www.tropos.com/pdf/\nchaska casestudy.pdf.\n[23]\nAnytime, Anywhere Broadband GoMoorhead!, Moorhead, Minnesota, A\nTropos Networks Case Study, November 2005, available at www.tropos.\ncom/pdf/casestudy gomoorehead.pdf.\n[24]\nMetro-Scale WiFi for Public Safety, San Mateo Police Department, A Tropos\nNetworks Case Study, March 2004, available at www.tropos.com/pdf/\nSMPD Casestudy.pdf.\n[25]\nMetro-Scale Video Surveillance: High-Profile Criminal Trial, A Tropos\nNetworks Case Study, August 2004, available at www.tropos.com/pdf/\npeterson casestudy.pdf.\n[26]\nJim Tomcik, MBFDD and MBTDD Wideband Mode: Technology Overview,\nhttp://grouper.ieee.org/groups/802/20/Contribs/C802.20-05-68r1.pdf.\n[27]\nPioneering Multi-Use Metro-Scale WiFi City of Corpus Christi, Texas, A\nTropos Networks Case Study, June 2005, available at www.tropos.com/\npdf/corpus casestudy.pdf.\n[28]\nNico Bayer, Dmitry Sivchenko, Bangnan Xu, Veselin Rakocevic, and\nJoachim Habermann, Transmission timing of signalling messages in IEEE\n802.16-based Mesh Networks, University of Applied Sciences, Giessen-\nFriedberg, Germany; Veselin Rakocevic, City University, UK, available at\nwww.staff.city.ac.uk/˜veselin/publications/Bayer EW2006.pdf.\n" }, { "page_number": 117, "text": "106\n■\nSecurity in Wireless Mesh Networks\n[29]\nSimone Redana, Matthias Lott, and Antonio Capone, Performance Evalua-\ntion of Point-to-Multi-Point (PMP) and Mesh Air-Interface in IEEE Standard\n802.16a, in Proceedings of IEEE VTC Fall 2004, Los Angeles, September\n2004.\n[30]\nFuqiang Liu, Zhihui Zeng, Jian Tao, Qing Li, and Zhangxi Lin, Achieving\nQoS for IEEE 802.16 in Mesh Mode, 8th International Conference on Com-\nputer Science and Informatics, Salt Lake City, Utah, July 21–26, 2005.\n[31]\nMika Kasslin and Dave Beyer, Mesh Mode Text for 802.16ab Standard,\nNokia, available at www.ieee802.org/16/tg4/contrib/802164c-01 39.pdf.\n[32]\nCity\nof\nTempe\nSelects\nStrix\nfor\nCity-wide\nWiFi\nMesh\nNetwork,\nCity of Tempe Deploying Mesh Networks, May 2006, available at\nwww.bbwexchange.com/ publications/page1386.asp.\n[33]\nStrix Systems Products and Solutions for Mesh Networks, June 2006, avail-\nable at www.strixsystems.com.\n[34]\nSimone Redana and Matthias Lott, Performance Analysis of IEEE 802.16a in\nMesh Operation Mode, in Proceedings of IST SUMMIT 2004, Lyon, France,\nJune 2004.\n[35]\nHung-Yu Wei, Samrat Ganguly, Rauf Izmailov, and Zygmunt J. Haas,\nInterference-Aware IEEE 802.16 WiMax Mesh Networks, in Proceedings\n61st IEEE Vehicular Technology Conference (VTC Spring ’05), 2005.\n[36]\nAlan Barry, George Healy, Cian Daly, Joseph Johnson, and Ronan J.\nSkehill, Overview of Wi-Max IEEE 802.16, in Proceedings of the 5th An-\nnual ICT Information Technology and Telecommunications, Cork, October\n2005.\n[37]\nD. Aguayo, J. Bicket, S. Biswas, D.S.J. De Couto, and R. Morris, MIT Roofnet\nImplementation, available at http://pdos.lcs.mit.edu/roofnet/design/.\n[38]\nMichael Carlberg Lax and Annelie Dammander, WiMAX — A Study of\nMobility and a MAC-layer Implementation in GloMoSim, Master’s thesis,\nUmea University, Sweden, April 2006.\n[39]\nNortel Solutions, Products and Experience in the Mesh Networking Field,\nJune 2006, available at www.nortelnetworks.com.\n[40]\nDave Beyer, Nico van Waes, and Carl Eklund, Tutorial: 802.16 MAC Layer\nMesh Extensions Overview, IEEE 802.16 Standard Group Discussions,\nFebruary 2002.\n[41]\nHarish Shetiya and Vinod Sharma, Algorithms for Routing and Centralized\nScheduling in IEEE 802.16 Mesh Networks, IEEE Wireless Communications\nand Networking Conference 2006, Las Vegas, April 2006.\n[42]\nProposal for Connection Oriented Mesh, available at www.ieee802.org/\n16/tgd/contrib/S80216d-03 18.pdf.\n[43]\nThikrait Al Mosawi, Review of Existing Mobile Broadband Wireless Access\n(MBWA) Technologies (IEEE 802.16 AND IEEE 802.20), Center for Telecom-\nmunications Research, November 2004.\n[44]\nIntroduction of the IEEE 802.20 Standard for Mobile Broadband Wireless\nAccess systems, RSAC paper 2/2004, available at www.ofta.gov.hk/en/ad-\ncomm/rsac/paper/rsac2-2004.pdf.\n[45]\nSystem Requirements for IEEE 802.20 Mobile Broadband Wireless Access\nSystems, available at www.ieee802.org/20/Contribs/C802.20-04-44.doc.\n" }, { "page_number": 118, "text": "Mesh Networking in Wireless PANs, LANs, MANs, and WANs\n■\n107\n[46]\nIEEE C802.20-03/104: IEEE 802.20 Working Group on Mobile Broad-\nband Wireless Access, available at www-sop.inria.fr/planete/qni/C802.20-\n03-104.pdf.\n[47]\n802.16 IEEE Standard for Local and Metropolitan Area Networks, Part 16:\nAir Interface for Fixed Broadband Wireless Access Systems, IEEE Computer\nSociety and IEEE Microwave Theory and Techniques Society.\n[48]\nStephen Roberts, Development of IEEE 802.16e Functionality in QualNet,\nMaster’s thesis, Dublin City University, School of Electronic Engineering,\nJanuary 2006.\n[49]\nD.B. Johnson, and D.A. Maltz, Dynamic source routing in ad-hoc wireless\nnetworks, In T. Imielinski and H. Korth, eds., Mobile Computing, Kluwer,\n1996, pp. 153–181.\n[50]\nK. Chandran, S. Raghunathan, and S.R. Prakash, A feedback-based scheme\nfor improving TCP performance in ad hoc wireless networks, IEEE Personal\nCommunications, 8, 1, 3439, 2001.\n[51]\nC. Perkins, E. Belding-Royer, and S. Das, Ad hoc On-demand Distance\nVector (AODV) Routing, IETF RFC 3561, July 2003.\n[52]\nD.S.J. De Couto, D. Aguayo, J. Bicket, and R. Morris, A high-throughput\npath metric for multi-hop wireless routing, in ACM Annual International\nConference on Mobile Computing and Networking (MOBICOM), 2003,\npp. 134–146.\n[53]\nE.M. Belding-Royer, Multi-level hierarchies for scalable ad hoc routing,\nACM/Kluwer Wireless Networks, 9, 5, 461–478, 2003.\n[54]\nA.K. Saha and D.B. Johnson, Self-organizing Hierarchical Routing for\nScalable Ad hoc Networking, Technical report, TR04-433, Department of\nComputer Science, Rice University.\n[55]\nK. Xu, X. Hong, and M. Gerla, Landmark routing in ad hoc networks with\nmobile backbones, in Proceedings of IEEE GLOBECOM 2000, San Francisco,\nNovember 2000, http://citeseer.ist.psu.edu/gerla00landmark.html.\n[56]\nC.-C. Shen, C. Srisathapornphat, and C. Jaikaeo, An adaptive management\narchitecture for ad hoc networks, IEEE Communications Magazine, 41, 2,\n108–115, 2003.\n[57]\nAoki et al., 802.11 TGs Simple Efficient Extensible, Mesh (SEE-Mesh) Pro-\nposal, IEEE 802 11-05/0562r01, 2005.\n[58]\nSheu et al., 802.11 TGs MAC Enhancement Proposal, IEEE 802 11-\n05/0575r4, 2005.\n[59]\nM. Papadopouli and H. Schulzrinne, Network Connection Sharing in an Ad\nhoc Wireless Network among Collaborative Hosts, NOSSDAV, June 24th–\n26th, 1999, Bell Labs, New Jersey.\n[60]\nQ. Xue and A. Ganz, Ad hoc QoS On-demand Routing (AQOR) Algorithm\nfor QoS Support in Mobile Ad hoc Networks, Technical report TR-CSE-\n02-09, J. Parallel Distrib. Comput., 63, 2, 154–165, 2003.\n[61]\nI.F. Akyildiz, J. Xie, and S. Mohanty, A survey of mobility management in\nnext-generation all-IP-based wireless systems, IEEE Wireless Communica-\ntions, 11, 4, 1628, 2004.\n[62]\nJ. Kajiya, Commodity Software Steerable Antennas for Mesh Networks,\nMicrosoft Mesh Networking Summit, June 2004.\n" }, { "page_number": 119, "text": "108\n■\nSecurity in Wireless Mesh Networks\n[63]\nT.-S. Yum and K.-W. Hung, Design algorithms for multihop packet radio\nnetworks with multiple directional antennas stations, IEEE Transactions on\nCommunications, vol. 40, pp. 1716–1724, November 1992.\n[64]\nA. Nasipuri, S. Ye, and R.E. Hiromoto, A MAC protocol for mobile ad hoc\nnetworks using directional antennas, in IEEE Wireless Communications and\nNetworking Conference (WCNC), 2000, pp. 1214–1219.\n[65]\nY.B. Ko, V. Shankarkumar, and N.H. Vaidya, Medium access control proto-\ncols using directional antennas in ad hoc networks, in IEEE Annual Con-\nference on Computer Communications (INFOCOM), 2000, pp. 1321.\n[66]\nR.R. Choudhury, X. Yang, R. Ramanathan, and N.H. Vaidya, Using direc-\ntional antennas for medium access control in ad hoc networks, in ACM\nAnnual International Conference on Mobile Computing and Networking\n(MOBICOM), 2002, pp. 5970.\n[67]\nX. Quin and R. Berry, Exploiting multiuser diversity for medium access\ncontrol in wireless networks, in Proceedings IEEE INFOCOM 2003, San\nFrancisco, 30 Mar.–3 Apr. 2003, vol. 2, pp. 108–494.\n[68]\nH. Jiang, W. Zhuang, X. (Sherman) Shen, A. Abdrabou, and P. Wang,\nDifferentiated services for wireless mesh backbone, Centre for Wireless\nCommunications (CWC), Department of Electrical and Computer Engineer-\ning, University of Waterloo, Canada, IEEE Communications Magazine, Vol-\nume 44, Issue 7, pp. 113–119, July 2006.\n[69]\nB. Karp and H. Kung, GPSR: Greedy perimeter stateless routing for wireless\nnetworks, in Proceedings ACM MOBICOM00, pp. 243–254.\n[70]\nS. C. Ergen, ZigBee/IEEE 802.15.4 Summary, September 10, 2004, available\nat http://www.cs.wisc.edu/∼suman/courses/838/readinglist.html.\n[71]\nG.R Hiertz, Y. Zang, S. Max, and H.-J. Reumerman, 802.15.5 MAC Design\nProposal, Philips Research Laboratories, ComNets, RWTH Aachen Univer-\nsity, http://mpa.comnets.rwth-aachen.de.\n[72]\nSupriya Maheshwari, An E cient QoS Scheduling Architecture for IEEE\n802.16 Wireless MANs. Master’s thesis, Indian Institute of Technology, Bom-\nbay, India, 2005.\n[73]\nIETF Integrated Services Working Group, available at www.ietf.org/html.\ncharters/IntServ-charter.htm.\n[74]\nIETF Differentiated Services Working Group, available at www.ietf.org/\nhtml.charters/DiffServ-charter.htm.\n" }, { "page_number": 120, "text": "SECURITY\nPROTOCOLS\nAND TECHNIQUES\nII\n" }, { "page_number": 121, "text": "" }, { "page_number": 122, "text": "Chapter 3\nAttacks and Security\nMechanisms\nAnjum Naveed, Salil S. Kanhere, and Sanjay K. Jha\nContents\n3.1\nIntroduction ......................................................... 112\n3.2\nSecurity Issues in Wireless Mesh Networks ........................ 114\n3.3\nAttacks in Wireless Mesh Networks ................................ 115\n3.3.1\nPhysical Layer Attacks ....................................... 115\n3.3.2\nMAC Layer Attacks ........................................... 115\n3.3.2.1\nPassive Eavesdropping ............................. 116\n3.3.2.2\nLink Layer Jamming Attack ........................ 116\n3.3.2.3\nMAC Spoofing Attack .............................. 116\n3.3.2.4\nReplay Attack ....................................... 117\n3.3.2.5\nPre-Computation and Partial Matching Attacks ... 118\n3.3.3\nNetwork Layer Attacks....................................... 119\n3.3.3.1\nControl Plane Attacks .............................. 119\n3.3.3.2\nData Plane Attacks ................................. 121\n3.3.4\nMulti-Radio Multi-Channel Wireless Mesh Network\nAttacks ........................................................ 122\n3.4\nCharacteristics of Security Solutions for Wireless\nMesh Networks...................................................... 125\n3.5\nSecurity Mechanisms for Wireless Mesh Networks................. 126\n3.5.1\nMAC Layer Security Mechanisms ............................ 127\n3.5.1.1\nIntrusion Prevention Mechanisms ................. 127\n3.5.1.2\nIntrusion Detection Mechanisms................... 130\n111\n" }, { "page_number": 123, "text": "112\n■\nSecurity in Wireless Mesh Networks\n3.5.2\nNetwork Layer Security Mechanisms........................ 130\n3.5.2.1\nIntrusion Prevention Mechanisms ................. 130\n3.5.2.2\nIntrusion Detection Mechanisms................... 131\n3.6\nToward Standardization ............................................. 132\n3.6.1\nVulnerabilities in IEEE 802.11i and Security Attacks........ 135\n3.6.1.1\nIEEE 802.1X Vulnerabilities ........................ 135\n3.6.1.2\nFour-Way Handshake Vulnerabilities .............. 137\n3.6.1.3\nCCMP Encryption Vulnerabilities .................. 139\n3.7\nOpen Issues ......................................................... 139\n3.8\nConclusion........................................................... 140\nReferences................................................................. 140\nThe true potential of any network cannot be exploited without consider-\ning and adequately addressing the security issues. Wireless mesh networks\n(WMNs), being multi-hop wireless networks, are prone to most of the secu-\nrity attacks on multi-hop wireless networks. In this chapter, we will discuss\nthe security vulnerabilities in multi-hop wireless networks that are relevant\nto WMNs. We will consider the attacks in WMNs and the possible solution\nmechanisms to prevent and counteract these attacks.\n3.1\nIntroduction\nIn recent years, WiFi (802.11) networks have become pervasive with nu-\nmerous hotspots being deployed in urban city centers. However, to be\nconnected, the mobile clients need to be within the radio range of the ac-\ncess point. To ensure that the target area is sufficiently covered, ISPs would\nneed to install additional hotspots in strategically placed locations to extend\nexisting coverage. This may not always be possible due to constraints on\nthe terrain, social issues, etc. Further, deploying additional hotspots adds to\nthe installation cost and more importantly to the running costs (subscription\ncost for Internet connectivity for each access point). A promising, low-cost\nalternative for providing last-mile wireless connectivity is the concept of\nWMNs, which are multi-hop wireless networks consisting of mesh routers\nand mesh clients. Generally, mesh routers have limited mobility and act\nas access points for the mobile clients to provide the connectivity over\nmultiple hops as well as route the traffic for neighboring mesh routers.\nSome of the routers are equipped with wired interface and serve the pur-\npose of gateway to provide the connectivity with the Internet. The clients’\nnodes may also act as intermediate hops for neighboring nodes to extend\nthe connectivity. A typical WMN architecture is shown in Figure 3.1. By en-\nabling multi-hop communication between the mesh nodes, it is possible for\n" }, { "page_number": 124, "text": "Attacks and Security Mechanisms\n■\n113\nG2\nRouter\nGateway\nClient mesh\nRouter/AP\nG1\nB\nE\nF\nH\nC\nClient\nInternet\nA\nD\nFigure 3.1\nWireless mesh network architecture.\nseveral mobile clients to share a single broadband connection to the Inter-\nnet. Several WMN deployments have been planned for major cities across\nthe globe (Taipei, Moscow, Philadelphia, etc.) in the near future. However,\nvery little attention has been devoted by the research community to address\nthe security issues in WMNs.\nThe broadcast nature of transmission and the dependency on the inter-\nmediate nodes for routing the user traffic leads to security vulnerabilities\nmaking WMNs prone to various attacks. The attacks can be external as well\nas internal in nature. External attacks are launched by intruders who are not\npart of the WMN and gain illegitimate access to the network. For example,\nan intruding node may eavesdrop on the packets and replay those packets\nat a later stage of time to gain access to the network resources. Attacks from\nexternal nodes can be prevented by resorting to cryptographic techniques\nsuch as encryption and authentication. On the other hand, the internal at-\ntacks are launched by the nodes that are part of the WMN. One example of\nsuch an attack is an intermediate node dropping the packets, which it was\nsupposed to forward, leading to a denial-of-service (DoS) attack. Similarly,\nthe intermediate node may keep the copy of all the data that it forwards\n(internal eavesdropping) for offline processing and meaningful informa-\ntion retrieval without the knowledge of any other node in the network.\nSuch attacks are typically launched either by selfish nodes or by malicious\nnodes, which may have been possibly compromised by attackers. There is\na subtle difference in their motives. The selfish node is seeking to greedily\nacquire greater than its fair share of the network resources at the expense\n" }, { "page_number": 125, "text": "114\n■\nSecurity in Wireless Mesh Networks\nof other users. On the contrary a malicious attacker’s sole aim is to un-\ndermine the performance of the entire network. Note that in an internal\nattack, the misbehaving node is part of the WMN and hence has access to\nall the keying and authentication information. Consequently, cooperative\nmechanisms, which enable other nodes within the network to detect and\npossibly isolate these misbehaving nodes, need to be employed.\nIt is evident that the true potential of WMN cannot be exploited without\nconsidering and adequately addressing the internal as well as the external\nsecurity issues. In this chapter, we identify the security issues in WMNs,\nfollowed by descriptions of attacks on WMNs. The primary focus will be\nthe attacks that affect the MAC layer and the network layer of WMNs.\nThe characteristics of the security solution for WMNs are identified and\ndifferent solution mechanisms are discussed. The standardization efforts\nfor the security in WMNs are discussed. The chapter is concluded with\nsome open issues yet to be considered in relation to security of WMNs.\n3.2\nSecurity Issues in Wireless Mesh Networks\nSeveral vulnerabilities exist in the protocols for WMNs that can be exploited\nby the attackers to degrade the performance of the network. The WMN\nnodes depend on the intermediate nodes for connectivity with other nodes\nin the network and the Internet. Consequently, the MAC layer protocols\nas well as the routing protocols for WMNs assume that the participating\nnodes are well behaved with no malicious intentions. Therefore, all the\nnodes are assumed to follow the MAC protocol and perform the routing\nand packet forwarding operations as specified by the respective protocols.\nBased on this assumed trust, the nodes make independent decisions for\ntheir transmission, depending on the wireless channel availability. Similarly,\nthe routing protocols require the WMN nodes to exchange their routing\ninformation within the neighborhood to make efficient routing decisions.\nBecause the nodes are assumed to be well behaved, each node makes\nan independent decision based on the routing protocol specifications. The\nnode then informs its neighbors about the decision. The neighbor nodes\nneither verify the decision nor the information transmitted by the node.\nIn practice, however, some WMN nodes may behave in a selfish manner\nand other nodes may be compromised by malicious users. The assumed\ntrust and the lack of accountability make the MAC layer protocols and the\nrouting protocols vulnerable to various active attacks, such as black hole\nattacks, wormhole attacks, and rushing attacks [11–13].\nThe malicious or selfish nodes can drop data packets selectively or\nmay choose to drop all the packets without forwarding any traffic. Further,\n" }, { "page_number": 126, "text": "Attacks and Security Mechanisms\n■\n115\nbecause the participating nodes may not be owned by one administrator,\nspecifically in case of community deployment of WMNs, data confidentiality\nand data integrity can be compromised if the intermediate node keeps the\ncopy of all the data for offline cryptanalysis and information retrieval. The\nmalicious nodes may also inject bad packets in the network, which may\nlead to a DoS attack. Similarly, passively sniffed packets can be replayed at\na later time to gain access to the network resources. All these vulnerabilities\nrender WMNs prone to security attacks. We consider the attacks on WMNs\nthat exploit these vulnerabilities in the next section.\n3.3\nAttacks in Wireless Mesh Networks\nIn this section, the details of various attacks on WMNs are given. We con-\nsider the attacks affecting the physical layer, MAC layer, and the network\nlayer because these layers form the core of the network. We do not consider\nthe attacks on the transport layer and the application layers because these\nlayers are primarily implemented in the end-user devices, hence the attacks\non these layers are independent of the underlying network. Therefore, the\nattacks and the counter-measures on these layers (application and trans-\nport) for WMNs, other wireless networks, or even wired networks would\nbe the same rather than being specific to WMNs.\n3.3.1\nPhysical Layer Attacks\nAll wireless networks, including WMNs, are vulnerable to radio jamming\nattacks at the physical layer. The radio jamming attack [14] is a potentially\ndamaging attack which can be launched with relative ease by simply allow-\ning a wireless device to transmit a strong signal, which can cause sufficient\ninterference to prevent packets in the victim network from being received.\nIn its simplest form, the attacker may continuously transmit the jamming\nsignal (constant jammer). Alternately, the attacker may resort to slightly so-\nphisticated strategies whereby the attacker only transmits the radio signal\nwhen it senses some activity on the channel and remains quiet otherwise\n(reactive jammer). However, these types of jamming attacks, where the\ntransmission is an arbitrary signal, can be regarded as noise in the channel\nand MAC protocols like BMAC [15] can successfully counteract these attacks\nto a certain degree by adjusting the signal-to-noise ratio (SNR) threshold\nat the receiving node. More complex forms of radio jamming attacks have\nbeen studied in [14], where the attacking devices do not obey the MAC layer\nprotocol. We discuss these attacks in Section 3.3.2 as link layer jamming\nattacks.\n" }, { "page_number": 127, "text": "116\n■\nSecurity in Wireless Mesh Networks\n3.3.2\nMAC Layer Attacks\n3.3.2.1\nPassive Eavesdropping\nThe broadcast nature of transmission of the wireless networks makes these\nnetworks prone to passive eavesdropping by the external attackers within\nthe transmission range of the communicating nodes. Multi-hop wireless\nnetworks like WMNs are also prone to internal eavesdropping by the inter-\nmediate hops, whereby a malicious intermediate node may keep the copy\nof all the data that it forwards, without knowledge of any other node in\nthe network. Although passive eavesdropping does not affect the network\nfunctionality directly, it leads to the compromise in data confidentiality and\ndata integrity. Data encryption is generally employed using strong encryp-\ntion keys to protect the confidentiality and integrity of data.\n3.3.2.2\nLink Layer Jamming Attack\nLink layer jamming attacks are more complex compared to blind physi-\ncal layer radio jamming attacks. Rather than transmitting random bits con-\nstantly, the attacker may transmit regular MAC frame headers (no payload)\non the transmission channel which conform to the MAC protocol being\nused in the victim network [16]. Consequently, the legitimate nodes always\nfind the channel busy and back off for a random period of time before sens-\ning the channel again. This leads to the denial of service for the legitimate\nnodes and also enables the jamming node to conserve its energy resources.\nIn addition to the MAC layer, jamming can also be used to exploit the net-\nwork and transport layer protocols [17]. Intelligent jamming is not a purely\ntransmit activity. Sophisticated sensors can be deployed, which detect and\nidentify victim network activity, with a particular focus on the semantics of\nhigher-layer protocols (e.g., AODV [Ad-hoc On-demand Distance Vector]\nand TCP). Based on the observations of the sensor, the attacker can exploit\nthe predictable timing behavior exhibited by higher-layer protocols and use\noffline analysis of packet sequences to maximize the potential gain for the\njammer. These attacks can be effective even if encryption techniques such\nas Wired Equivalent Privacy (WEP) and WiFi Protected Access (WPA) have\nbeen employed. This is because the sensor that assists the jammer can\nstill monitor the packet size, timing, and sequence to guide the jammer.\nBecause these attacks are based on carefully exploiting protocol patterns\nand consistencies across size, timing, and sequence, preventing them will\nrequire modifications to the protocol semantics so that these consistencies\nare removed wherever possible.\n3.3.2.3\nMAC Spoofing Attack\nMAC addresses have long been used as the singularly unique layer-2 net-\nwork identifiers in both wired and wireless LANs. MAC addresses which\n" }, { "page_number": 128, "text": "Attacks and Security Mechanisms\n■\n117\nare globally unique have often been used as an authentication factor or as\na unique identifier for granting varying levels of network privileges to a\nuser. This is particularly common in 802.11 WiFi networks. However, to-\nday’s MAC protocols (802.11) and network interface cards do not provide\nfor any safeguards that would prevent a potential attacker from modifying\nthe source MAC address in its transmitted frames. On the contrary, there is\noften full support in the form of drivers from manufacturers, which makes\nthis particularly easy. Modifying the MAC address in transmitted frames is\nreferred to as MAC spoofing, and can be used by attackers in a variety\nof ways. MAC spoofing enables the attacker to evade Intrusion Detection\nSystems (IDSs) that are in place. Further, today’s network administrators of-\nten use MAC addresses in access control lists. For example, only registered\nMAC addresses are allowed to connect to the access points. An attacker\ncan easily eavesdrop on the network to determine the MAC addresses of\nlegitimate devices. This enables the attacker to masquerade as a legitimate\nuser and gain access to the network. An attacker can even inject a large\nnumber of bogus frames into the network to deplete the resources (in par-\nticular, bandwidth and energy), which may lead to denial of service for the\nlegitimate nodes.\n3.3.2.4\nReplay Attack\nThe replay attack, often known as the man-in-the-middle attack [18], can\nbe launched by external as well as internal nodes. An external malicious\nnode (not part of WMN) can eavesdrop on the broadcast communication\nbetween two nodes (A and B) in the network, as shown in Figure 3.2. It\nNode A\nNode B\nAdversary\nData−1\nData−2\nData−3\nData−4\nData−3\nReplayed/MAC spoofed\nMay grant adversary\nwith unauthorized\naccess\nFigure 3.2\nRobustness against MAC spoofing and replay attacks.\n" }, { "page_number": 129, "text": "118\n■\nSecurity in Wireless Mesh Networks\ncan then transmit these legitimate messages at a later stage of time to gain\naccess to the network resources. Generally, the authentication information\nis replayed where the attacker deceives a node (node B in Figure 3.2) to\nbelieve that the attacker is a legitimate node (node A in Figure 3.2). On\na similar note, an internal malicious node, which is an intermediate hop\nbetween two communicating nodes, can keep a copy of all relayed data.\nIt can then retransmit this data at a later point in time to gain the unau-\nthorized access to the network resources. The replay attack, exploiting the\nIEEE 802.1X [33] authentication mechanism, is discussed in Section 3.6.\n3.3.2.5\nPre-Computation and Partial Matching Attacks\nIn this section we discuss a different form of security attacks. Unlike the\nabove-mentioned attacks where MAC protocol vulnerabilities are exploited,\nthese attacks exploit the vulnerabilities in the security mechanisms that are\nemployed to secure the MAC layer of the network. Pre-computation and\npartial matching attacks exploit the cryptographic primitives that are used\nat MAC layer to secure the communication. In a pre-computation attack\nor Time Memory Trade-Off attack (TMTO), the attacker computes a large\namount of information (key, plaintext, and respective cipher text) and stores\nthat information before launching the attack. When the actual transmission\nstarts, the attacker uses the pre-computed information to speed up the\ncryptanalysis process. TMTO attacks are highly effective against a large\nnumber of cryptographic solutions. On the other hand, in a partial matching\nattack, the attacker has access to some (cipher text, plaintext) pairs, which\nin turn decreases the encryption key strength and improves the chances of\nsuccess of the brute force mechanisms. Partial matching attacks exploit the\nweak implementations of encryption algorithms. For example, in the IEEE\n802.11i standard for MAC layer security in wireless networks [30], the MAC\naddress fields in the MAC header are used in the message integrity code\n(MIC). The MAC header is transmitted as plaintext while the MIC field is\ntransmitted in the encrypted form. Partial knowledge of the plaintext (MAC\naddress) and the cipher text (MIC) makes IEEE 802.11i vulnerable to partial\nmatching attacks.\nDoS attacks may also be launched by exploiting the security mecha-\nnisms. For example, the IEEE 802.11i standard for MAC layer security in\nwireless networks is prone to the session hijacking attack and the man-\nin-the-middle attack, exploiting vulnerabilities in IEEE 802.1X, and DoS\nattack, exploiting vulnerabilities in the four-way handshake procedure in\nIEEE 802.11i. Although these attacks are also considered as MAC layer at-\ntacks, we pend the discussion on IEEE 802.11i, its vulnerabilities, attacks\nexploiting these vulnerabilities, and the proposed prevention mechanisms\ntill Section 3.6.\n" }, { "page_number": 130, "text": "Attacks and Security Mechanisms\n■\n119\n3.3.3\nNetwork Layer Attacks\nThe attacks on the network layer can be divided into control plane attacks\nand data plane attacks and can be active or passive in nature. Control plane\nattacks generally target the routing functionality of the network layer. The\nobjective of the attacker is to make routes unavailable or force the network\nto choose sub-optimum routes. On the other hand, the data plane attacks\naffect the packet forwarding functionality of the network. The objective\nof the attacker is to cause the denial of service for the legitimate user by\nmaking user data undeliverable or injecting malicious data into the network.\nWe first consider the network layer control plane attacks, followed by a\ndiscussion on network layer data plane attacks.\n3.3.3.1\nControl Plane Attacks\nRushing attacks [11] targeting the on-demand routing protocols (e.g., AODV)\nwere among the first exposed attacks on the network layer of multi-hop\nwireless networks. Rushing attacks exploit the route discovery mechanism\nof on-demand routing protocols. In these protocols, the node requiring the\nroute to the destination floods the Route Request message, which is identi-\nfied by a sequence number. To limit the flooding, each node only forwards\nthe first message that it receives and drops remaining messages with the\nsame sequence number. The protocols specify a specific amount of delay\nbetween receiving the Route Request message by a particular node and\nforwarding it, to avoid collusion of these messages. The malicious node\nlaunching the rushing attack forwards the Route Request message to the\ntarget node before any other intermediate node from source to destination.\nThis can easily be achieved by ignoring the specified delay. Consequently,\nthe route from source to destination includes the malicious node as an in-\ntermediate hop, which can then drop the packets of the flow resulting in\ndata plane DoS attack.\nA wormhole attack has a similar objective albeit it uses a different tech-\nnique [12]. During a wormhole attack, two or more malicious nodes col-\nlude together by establishing a tunnel using an efficient communication\nmedium (i.e., wired connection or high-speed wireless connection, etc.),\nas shown in Figure 3.3. During the route discovery phase of on-demand\nrouting protocols, the Route Request messages are forwarded between the\nmalicious nodes using the established tunnel. Therefore, the first Route Re-\nquest message that reaches the destination node is the one forwarded by\nthe malicious nodes. Consequently, the malicious nodes are added in the\npath from source to destination. Once the malicious nodes are included in\nthe routing path, the malicious nodes either drop all the packets, result-\ning in complete denial of service, or drop the packets selectively to avoid\ndetection.\n" }, { "page_number": 131, "text": "120\n■\nSecurity in Wireless Mesh Networks\nS\nM1\nM2\nD\nRReq\nData\ndropped\nRReq\nSource\nDestination\nRReq\nRReply\nTunnel\nMalicious nodes\nFigure 3.3\nWormhole attack launched by nodes M1 and M2. Nodes use high-speed\ntunnel to forward routing protocol control messages while data is dropped.\nA black hole attack (or sink hole attack) [19] is another attack that leads\nto denial of service in wireless mesh networks. It also exploits the route dis-\ncovery mechanism of on-demand routing protocols. In a black hole attack,\nthe malicious node always replies positively to a Route Request although it\nmay not have a valid route to the destination. Because the malicious node\ndoes not check its routing entries, it will always be the first to reply to the\nRoute Request message. Therefore, almost all the traffic within the neigh-\nborhood of the malicious node will be directed toward the malicious node,\nwhich may drop all the packets, resulting in denial of service. Figure 3.4\nshows the effect of a black hole attack in the neighborhood of the mali-\ncious node where all the traffic is directed toward the malicious node. A\nmore complex form of the attack is the cooperative black hole attack where\nmultiple malicious nodes collude together, resulting in complete disruption\nM\nM replies positively to every route request\nData dropped\nData\nFigure 3.4\nBlack hole attack. Node M replies positively to every Route Request.\nConsequently all data is forwarded to the node, which then drops the data.\n" }, { "page_number": 132, "text": "Attacks and Security Mechanisms\n■\n121\nof routing and packet forwarding functionality of the network. The coopera-\ntive black hole attack and the prevention mechanisms have been studied\nin [13].\nA grey hole attack is a variant of the black hole attack. In a black\nhole attack, the malicious node drops all the traffic that it is supposed to\nforward. This may lead to possible detection of the malicious node. In a\ngrey hole attack, the adversary avoids the detection by dropping the packets\nselectively. A grey hole attack does not lead to complete denial of service,\nbut it may go undetected for a longer duration of time. This is because the\nmalicious packet dropping may be considered congestion in the network,\nwhich also leads to selective packet loss.\nA Sybil attack is the form of attack where a malicious node creates mul-\ntiple identities in the network, each appearing as a legitimate node [20]. A\nSybil attack was first exposed in distributed computing applications where\nthe redundancy in the system was exploited by creating multiple identi-\nties and controlling the considerable system resources. In the networking\nscenario, a number of services like packet forwarding, routing, and col-\nlaborative security mechanisms can be disrupted by the adversary using\na Sybil attack. Following form of the attack affects the network layer of\nWMNs, which are supposed to take advantage of the path diversity in the\nnetwork to increase the available bandwidth and reliability. If the mali-\ncious node creates multiple identities in the network, the legitimate nodes,\nassuming these identities to be distinct network nodes, will add these iden-\ntities in the list of distinct paths available to a particular destination. When\nthe packets are forwarded to these fake nodes, the malicious node that cre-\nated the identities processes these packets. Consequently, all the distinct\nrouting paths will pass through the malicious node. The malicious node\nmay then launch any of the above-mentioned attacks. Even if no other at-\ntack is launched, the advantage of path diversity is diminished, resulting in\ndegraded performance.\nIn addition to the above-mentioned attacks, the wireless mesh networks\nare also prone to network partitioning attacks and routing loop attacks.\nIn a network partitioning attack, the malicious nodes collude together to\ndisrupt the routing tables in such a way that the network is divided into\nnon-connected partitions, resulting in denial of service for a certain network\nportion. Routing loop attacks affect the packet-forwarding capability of the\nnetwork where the packets keep circulating in loop until they reach the\nmaximum hop count, at which stage the packets are simply discarded.\n3.3.3.2\nData Plane Attacks\nData plane attacks are primarily launched by the selfish and malicious (com-\npromised) nodes in the network and lead to performance degradation or\ndenial of service for the legitimate user data traffic. The simplest of the\n" }, { "page_number": 133, "text": "122\n■\nSecurity in Wireless Mesh Networks\ndata plane attacks is passive eavesdropping. Eavesdropping has already\nbeen discussed in Section 3.3.2 as a MAC layer attack and we do not dis-\ncuss it further. Selfish behavior of the participating WMN nodes is a major\nsecurity issue because the WMN nodes are dependent on each other for\ndata forwarding. The intermediate-hop selfish nodes may not perform the\npacket-forwarding functionality as per the protocol. The selfish node may\ndrop all the data packets, resulting in complete denial of service, or it\nmay drop the data packets selectively or randomly. It is hard to distinguish\nbetween such a selfish behavior and the link failure or network conges-\ntion. On the other hand, malicious intermediate-hop nodes may inject junk\npackets into the network. Considerable network resources (bandwidth and\npacket processing time) may be consumed to forward the junk packets,\nwhich may lead to denial of service for the legitimate user traffic. The mali-\ncious nodes may also inject the maliciously crafted control packets, which\nmay lead to the disruption of routing functionality. The control plane attacks\nare dependent on such maliciously crafted control packets. The malicious\nand selfish behavior has been studied in [22,23].\n3.3.4\nMulti-Radio Multi-Channel Wireless Mesh\nNetwork Attacks\nIn this section, we consider the attacks that affect the network layer as\nwell as the MAC layer of WMNs. These attacks exploit the channel assign-\nment and routing algorithms in multi-radio multi-channel wireless mesh\nnetworks (MR-MC WMN). Bandwidth capacity is a major limitation for wire-\nless mesh networks. In MR-MC WMN, each WMN node is equipped with\nmultiple radios to increase the available bandwidth. Orthogonal channels\nare used for each interface of the node, which ensures simultaneous com-\nmunication using all the wireless interfaces without interference. Dynamic\nchannel assignment is required to assign the channels to the network links.\nThe objective of the channel assignment algorithms is to ensure the mini-\nmum interference within a WMN. Various joint routing and channel assign-\nment algorithms have been proposed for MR-MC WMN [1–5]. Readers are\nencouraged to review the dynamic routing and channel assignment algo-\nrithms proposed in [2] for better understanding of the attacks discussed in\nthis section. Note that channel assignment is done at the MAC layer while\nthe routing is a network layer functionality. All the joint routing and chan-\nnel assignment algorithms assume that the mesh nodes are well-behaved.\nHence the nodes make independent decisions about their channel assign-\nment based on the neigbhor channel assignment information and inform\nneighboring nodes about the decision, which is not verified. The assumed\n" }, { "page_number": 134, "text": "Attacks and Security Mechanisms\n■\n123\nInternet\nG2\nG1\nB\nE\nF\nC\nH\nI\nMalicious\nnode\nInterfering links\nAffected links\nA\nD\nFigure 3.5\nNetwork endo-parasite attack (NEPA). Assuming the node F is within\ninterference domain of node G.\ntrust among the WMN nodes and the independent decision of the nodes\nmake these algorithms vulnerable to security attacks.\nA network endo-parasite attack (NEPA) [21] is launched by the com-\npromised malicious node when it changes the channel assignment of its\ninterfaces in such a way that the interference on heavily loaded high prior-\nity channels increases (each interface is switched to a different high-priority\nchannel). This is contrary to the normal operation of the channel assign-\nment algorithm where the node assigns the least loaded channels to its\ninterfaces. Figure 3.5 shows the attack. The malicious node F has switched\nthe channel on link FH to the same channel as the link GC and link FI\nto the channel used by link GD. The malicious switching by node F will\nincrease the interference on links GC and GD. The malicious node does\nnot inform its neighbors about the change in channel assignment; there-\nfore, the neighboring nodes are unable to adjust their channel assignment\nto mitigate the effect of increased interference. The increase in interference\nresults in serious performance degradation.\nA channel ecto-parasite attack (CEPA) [21] is a special type of NEPA. Dur-\ning CEPA, the malicious node switches all its interfaces to the most heavily\nloaded highest priority channel. Like NEPA, the malicious node does not\ninform its interference domain neighbors about the change in channel as-\nsignment. The effect of the attack is the hidden usage of the most heavily\nloaded channel, which increases the interference considerably, resulting in\na decrease in performance. The attack is shown in Figure 3.6 where the\nmalicious node has switched both its child links FH and FI to the channel\n" }, { "page_number": 135, "text": "124\n■\nSecurity in Wireless Mesh Networks\nInternet\nG2\nG1\nA\nD\nB\nE\nF\nC\nH\nI\nAffected link\nMalicious\nnode\nInterfering links\nFigure 3.6\nChannel ecto-parasite attack (CEPA) (assuming the node F is within\ninterference domain of node G).\nthat is being used by the high-priority link GC. As the links FH and FI are\nwithin the interference range of the link GC, the link GC will experience\nhigh interference. However, the malicious node has not informed its neigh-\nbors about the change in channel assignment; therefore, the node G will\ncontinue to use the same channel on link GC, assuming the external noise\nor other factors to be the reason for degraded performance.\nA low cost ripple effect attack (LORA) [21] is launched when the compro-\nmised malicious node transmits misleading channel assignment information\nabout its interfaces to the neighboring nodes without actually changing the\nchannel assignment. The information is calculated in such a way that the\nneighboring nodes are forced to adjust their channel assignments to mini-\nmize the interference, which may generate a series of changes even in the\nchannel assignment of the nodes that are not direct neighbors of the mali-\ncious node. The effect of the attack is shown in Figure 3.7 using the arrow.\nAlthough most of the dynamic channel assignment algorithms prevent the\nripple effect to propagate within the network from the parent nodes (closer\nto the wired gateway) to the child nodes, the effect can still propagate in\nthe reverse direction. The objective of the attack is to force the network in\nthe quasi-stable state by imposing premature channel adjustment on other\nnodes repeatedly. Considerable network resources are consumed for chan-\nnel adjustment and the user data forwarding capability is severely affected.\nThe attack is relatively more severe than NEPA and CEPA because the effect\nis propagated to a large portion of the network even beyond the neighbors\nof the compromised node, disrupting the traffic forwarding capability of\nvarious nodes for considerable time duration.\n" }, { "page_number": 136, "text": "Attacks and Security Mechanisms\n■\n125\n250(l)\n500(i)\n500(j)\n250(j)\n250(m)\n125(k)\n250(k)\n250(n)\n500(m)\n500(n)\n250(l)\n250(i)\n250(m)\n250(k) 250(l)\n250(i)\n250(k)\n125(l)\n125(j)\n125(m)\n125(n)\n125(j)\n125(l)\n250(i)\n250(k)\n125(i)\n125(j)\n125(i)\nG1\nG2\nA\nD\nB\nC\nE\nF\nH\nI\nJ\nM\nFigure 3.7\nExample WMN with routers physically arranged in grid topology. G1\nand G2 are gateways connected to wired network. Edges show routing topology\nand labels along edges are bandwidth in kbps (channel). For simplicity, (k+ 1 )-hop\nneighbors include immediate physical neighbors only. Arrows show propagation of\nripple effect attack from compromised node M.\n3.4\nCharacteristics of Security Solutions for Wireless\nMesh Networks\nIn the previous section, we discussed the security attacks that exploit the\nvulnerabilities in the MAC layer and the network layer protocols for WMN.\nWe now list the essential characteristics that a security mechanism for WMN\nshould have to successfully prevent, detect, and counter these attacks. We\nonly list the characteristics that differentiate WMN security mechanisms from\nexisting security mechanisms for wired and wireless networks.\n■\nIn wired networks, the security services of data confidentiality and\ndata integrity are generally provided on a per-link basis (between\ntwo devices). This is based on the assumption that the end devices\nare secure. However, as discussed in previous sections, the WMN\nnodes may resort to the selfish and malicious behavior. To coun-\nteract the selfish and malicious behavior of the intermediate-hop\nnodes, the WMN must provide the end-to-end services of data con-\nfidentiality and data integrity, in addition to the security services on\na per-link basis.\n" }, { "page_number": 137, "text": "126\n■\nSecurity in Wireless Mesh Networks\n■\nThe trust establishment mechanism should be robust against inter-\nnal selfish and malicious behavior. Note that the internal selfish and\nmalicious nodes are part of WMNs, therefore the conventional au-\nthentication mechanisms based on cryptographic primitives may not\nbe effective against the internal misbehavior.\n■\nSection 3.3.3 and Section 3.3.4 indicate that the accountability should\nbe a necessary characteristic for WMNs to ensure that the WMN\nnodes behave according to the protocol specification even if the\nnodes make independent decisions about routing and channel\nassignment.\n■\nWireless mesh networks are self-administered networks and lack the\ncentralized administration authority which can respond to the net-\nwork issues. Therefore, the attack and anomaly detection mecha-\nnisms for wireless mesh networks should be self-sufficient and must\nnot be dependent on the administrator to verify the possible attack\nand anomaly alerts.\n■\nAn important characteristic of wireless mesh networks is the self-\nhealing nature. Therefore, the detection mechanisms must be cou-\npled with adequate automated response to the security attacks and\nidentified anomalies.\nHaving identified the essential characteristics of the security mechanisms for\nwireless mesh networks, we now consider different security mechanisms\nthat are employed to counter the attacks identified in Section 3.3.\n3.5\nSecurity Mechanisms for Wireless\nMesh Networks\nITU-T Recommendation X.800 [29]—Security Architecture for OSI—defines\nthe required security services for communication networks. The security\nservices have been broadly categorized into five groups: authentication, ac-\ncess control or authorization, confidentiality, integrity, and non-repudiation.\nSecurity management services have also been defined aimed at ensuring\navailability, accountability, and event management. The security services\ncan be categorized into two broad categories: intrusion prevention and in-\ntrusion detection. In case of intrusion prevention, measures are taken to\nstop the attacker from intruding into the network and launching the attack\non the network. The protection can be from external as well as internal\nintruders. Security services of authentication, access control, data confi-\ndentiality, data integrity, and non-repudiation lead to intrusion prevention.\nHowever, intrusion prevention is insufficient to protect the network from all\nattacks because no prevention technique can ensure complete protection.\n" }, { "page_number": 138, "text": "Attacks and Security Mechanisms\n■\n127\nAvailability,\naccountability\nIntrusion\ndetection system\nAutomated\nresponse\nIntrusion detection and\nautomated response\nSecure routing &\nchannel assignment\nAuthentication\nData integrity\nData confidentiality\nAuthorization\nNetwork layer\nIntrusion prevention\nNetwork layer\nMAC layer\nMAC layer\nFigure 3.8\nSecurity model for wireless mesh networks.\nTherefore, the intrusion prevention mechanisms are complemented by in-\ntrusion detection and response mechanisms. The role of intrusion detection\nis to identify the illegitimate activities which may be the consequence of the\nattacks or may lead to the attacks. Early detection and timely response can\nlimit the effect of the attack on the network. The intrusion detection and\nresponse mechanisms aim at ensuring the accountability and availability of\nthe network services. Figure 3.8 shows how different security services fit\ntogether in the security model for wireless mesh networks. We now con-\nsider the intrusion prevention mechanisms as well as intrusion detection\nmechanisms both at the MAC layer and the network layer of wireless mesh\nnetworks.\n3.5.1\nMAC Layer Security Mechanisms\n3.5.1.1\nIntrusion Prevention Mechanisms\nVarious security frameworks [30–32] have been proposed for multi-hop\nwireless networks that are applicable to wireless mesh networks with slight\nmodification. These security frameworks provide the security services of\nauthentication, data confidentiality, and data integrity at the MAC layer of\nthe network on a per-link basis. Most of the security frameworks employ\nthe cryptographic primitives. For example, Soliman and Omari [31] have\nproposed the security framework based on stream cipher for encryption to\nprovide the services of data confidentiality, data integrity, and authentica-\ntion. The objective of using stream cipher is to allow the online processing\nof the data. Consequently, minimum delay is introduced because of the\nsecurity provisioning. Two secret security keys, Secret Authentication Key\n(SAK) and Secret Session Key (SSK), are used for authentication of the\n" }, { "page_number": 139, "text": "128\n■\nSecurity in Wireless Mesh Networks\nsupplicant and authenticator. SAK is exchanged between the supplicant\nand the authenticator after initial mutual authentication from the authenti-\ncation server, whereas the SSK is used for a given communication session\nbetween the two nodes. The SAK and SSK pair is used by the communicat-\ning nodes to generate the permutation vector (PV), which is used for the\nencryption and decryption of data. In the strongest mode of security, the\ndata is also involved in the PV generation. The synchronization of the gen-\nerated permutation vector between the sender and the receiver of the data\nresults in origin authentication of every MAC Protocol Data Unit (MPDU).\nTo minimize the security overhead, plaintext MPDU is XORed with the PV\ngenerated for that MPDU. The authors have proved that the encryption of\ndata using PV provides strong security services of data confidentiality, data\nintegrity, and origin authentication.\nIEEE 802.11i was ratified in June 2004 as the standard for the security\nof the MAC layer of the wireless networks. The standard is based on the\ncryptographic primitives and provides the services of data confidentiality,\ndata integrity, and authentication. The standard is discussed in detail in\nSection 3.6.\nOne of the major security requirements in case of multi-hop wireless net-\nworks like WMN is the trust establishment between communicating nodes.\nAs mentioned in Section 3.4, conventional cryptography-based mechanisms\nare generally non-applicable to multi-hop networks like WMN. Conse-\nquently, a number of distributed neighbor-collaboration authentication pro-\ntocols have been proposed by researchers for this purpose [38,39,42]. A\ncomprehensive analysis of the authentication protocols for wireless net-\nworks can be found in [41]. Deng et al. [42] have proposed the threshold\nand identity-based authentication and key management for multi-hop wire-\nless networks. A threshold cryptography-based solution is proposed for the\ndistribution of the master key and the authen-\ntication of the nodes based on the private key. In the proposed scheme,\nall nodes possess the public key while every node has got a share of the\nprivate key. (k,n) Threshold secret sharing is employed to generate the pri-\nvate key for the node which states that “k” out of “n” shares of private key\nare required to construct the complete private key and less than k shares\nof the secret key cannot construct the complete private key. Based on this\nmechanism, whenever a node needs to refresh its private key, it needs k\nneighbors to send their secret share to the node to reconstruct the private\nkey and no node can construct the private key based on its own informa-\ntion. The process of private key generation is shown in Figure 3.9, where\nthe requesting node broadcasts the request message along with its own\nshare for verification. The neighboring nodes reply to the request message\nby sending their own share of the secret key to the requesting node. The\nrequesting node is able to generate the private key on receiving k shares\nof the key. Using this mechanism, the intruding node cannot generate the\n" }, { "page_number": 140, "text": "Attacks and Security Mechanisms\n■\n129\nA\nBroadcast own\nprivate key share\nReply from k\nneighbors with\ntheir private key share\nA is dependent on K neighbors\nfor complete private key generation\nFigure 3.9\nNeighbor collaboration for private key generation in wireless mesh\nnetworks.\nprivate key unless its own share of private key is verified by k neighboring\nnodes. Similarly, the private key of the misbehaving node is not refreshed\nby the neighbors. Therefore, the threshold secret sharing serves as the\nstrong authentication and key management solution.\nThe security mechanisms discussed above prevent the network from\nMAC layer attacks as follows. The security service of data confidentiality\nleads to the protection against passive eavesdropping attack. Although the\nnodes within the transmission range of the communicating nodes can still\noverhear the communication, the data is protected using encryption mecha-\nnisms provided by the data confidentiality service. Therefore, the received\ninformation is useless, unless it is decrypted using brute force methods,\nwhich are impractical, keeping in view the value of information retrieved\nversus the cost of attack. Data and header integrity service provides the\nprotection against MAC spoofing attacks. The message with spoofed MAC\naddress (IP address for IP spoofing) will fail the integrity check at the re-\nceiving node and will be discarded. Per-packet authentication and integrity\nprovided by the solutions [30,31] protect the data against replay attacks.\nThese solutions use a fresh key for each message which is synchronously\ncomputed by the sender and the receiver. Therefore, a replayed packet,\nencrypted using an outdated key, will fail the integrity check and will be\ndiscarded. Use of a fresh key for each message also protects the data from\npre-computation and partial matching attacks because the pre-computed\ninformation needs to be applied on every message to decrypt that mes-\nsage. This renders the attack extremely costly compared to the information\nretrieved.\n" }, { "page_number": 141, "text": "130\n■\nSecurity in Wireless Mesh Networks\n3.5.1.2\nIntrusion Detection Mechanisms\nVery few intrusion detection systems have been proposed at the MAC layer\nof wireless networks. Lim et al. [43] have proposed an intrusion detection\nsystem to secure wireless access points coupled with automated active re-\nsponse. The authors have proposed the deployment of specific detection\ndevices closer to wireless access points and the detection is done at the MAC\nlayer. RTS/CTS (Ready To Send/Clear To Send) messages from the black-\nlisted MAC addresses are proposed as detection metrics. As a response to\nthe intrusion, the authors propose the use of the intruder’s tactics back onto\nthe intruder by crafting and transmitting the malformed packets back. The\nproposed idea of deploying dedicated detection devices may not be cost\neffective. Similarly, the response mechanism may be computation resource\nextensive. Further, the legitimate nodes may get punished if the detected\ninformation is not accurate.\nOne of the most recent works in this context is from Liu et al. [24]. The\nauthors have proposed the game theoretic approach for selecting the opti-\nmum intrusion detection strategy at a given instance from a set of deployed\nweak intrusion detection mechanisms. The basic idea is that different in-\ntrusion detection techniques are very good at detecting certain types of\nattacks, but do not perform optimally in other cases. The combination of\nthese strategies and the use of optimum strategy in a given scenario can\nincrease the detection accuracy of the resulting system. However, while the\nidea of selecting the optimum technique at a given instance has strength,\nbasically at a given instance of time, only one weak intrusion detection\ntechnique will be used. Consequently, the performance of intrusion detec-\ntion may not significantly improve as compared to the increase in overhead\nbecause of the IDS selection mechanism.\nThe intrusion detection mechanisms at the MAC layer are used to detect\nthe attacks launched by misbehaving nodes that do not obey the MAC\nlayer protocol. These attacks include the link layer jamming attacks and\nDoS attacks.\n3.5.2\nNetwork Layer Security Mechanisms\n3.5.2.1\nIntrusion Prevention Mechanisms\nIntrusion prevention techniques have been proposed to secure the rout-\ning protocols for multi-hop wireless networks. These protocols include Se-\ncure Routing Protocol (SRP) [6], Secure AODV (SAODV) [7], Authenticated\nRouting for Ad hoc Network (ARAN) [8] and Ariadne, a secure on-demand\nrouting protocol [9], to list a few. The most recent work in this domain is\ndescribed in [10]. All these protocols use cryptographic primitives to estab-\nlish some form of trust between the network nodes through the process of\nmutual authentication. For example, SRP [6] is aimed at securing the route\n" }, { "page_number": 142, "text": "Attacks and Security Mechanisms\n■\n131\ndiscovery process and safeguards the routing functionality from attacks ex-\nploiting the routing protocol itself. The Route Request and Route Reply\nmessages are protected by message authentication code (MAC) for authen-\ntication of the originating node. The IP address of the intermediate nodes\nis also added in the Route Request message for cross validation to prevent\nthe network from black hole and wormhole attacks. The authors prove that\nthe protection of Route Request and Route Reply messages ensures prote-\nction against multiple attacks except for the case where multiple nodes\ncollude together and launch the attack. SAODV [7] uses digital signatures\nto authenticate all the fields of Route Request and Route Reply messages\nexcept from the hop count field. Digital signatures are used on end-to-end\nbasis between source and destination. The hop count field is secured using\nhash-chains on per-link basis.\nThe intrusion prevention mechanisms are primarily used to establish the\ntrust between the participating nodes and providing the control message\nintegrity and confidentiality. These services can provide some protection\nagainst wormhole and black hole attacks. However, the problem of mali-\ncious and misbehaving nodes cannot be addressed completely using the\nintrusion prevention mechanisms at the network layer and the support from\nintrusion detection mechanisms becomes mandatory.\n3.5.2.2\nIntrusion Detection Mechanisms\nNumerous intrusion detection techniques have been proposed at the net-\nwork layer for wired as well as wireless networks. In this section we briefly\ndiscuss some of the recent research efforts in this domain; however, the\nsurvey by no means is exhaustive. Most of the intrusion detection systems\nrely on the knowledge-based systems and data mining techniques [25–28].\nFor example, Huang et al. [26] have proposed IDS for multi-hop mobile\nwireless networks based on the cross-feature analysis. The nodes monitor\ndifferent parameters in the network and, based on values of (i −1) param-\neters, predict the value of ith parameter and compare it with monitored\nvalue of that parameter to detect routing anomalies in the networks. The\nauthors have also proposed the distributed cluster-based approach as an\nextension to this work [27], where they propose the division of networks\ninto clusters and only few elected nodes within each cluster perform the\nmonitoring with the intrusion detection probability almost the same as with\nall the nodes actively monitoring. This scheme is resource efficient, which\nis the primary design goal for wireless networks.\nYang et al. [28] have proposed the self-organized network layer se-\ncurity solution for mobile ad hoc networks. This is one of the very few\nsolutions which ensure self-healing and self-organized networks. The solu-\ntion is based on distributed neighbor collaboration and information cross-\nvalidation, resulting in self-organized, self-healing networks. The scheme\nis based on the threshold secret sharing discussed in Section 3.5.1 which is\n" }, { "page_number": 143, "text": "132\n■\nSecurity in Wireless Mesh Networks\nused to refresh the token of the nodes. The authors have proposed a novel\ntoken-based crediting scheme. The token of the node expires after a specific\ntime duration. The token expiry time of the node depends upon the credit\nof the node. The credit of the well-behaving nodes gets accumulated over\nthe period of time. Therefore, the token expiry time of these nodes is longer\nand is linearly incremented every time the node refreshes its token. The\ntoken of malicious or selfish nodes is revoked by neighbor collaboration\nrefraining them to participate in the network. The detection metrics used\nto differentiate between well-behaving and malicious nodes are based on\nthe routing protocols and consist of hop count distance, packet forwarding\nratio, etc.\nThe intrusion detection mechanisms at the network layer primarily ad-\ndress the issues of malicious, selfish, and misbehaving nodes that are at\nthe heart of almost all the attacks at the network layer. The solutions de-\nscribed in [26–28] identify the anomalies in the control messages to detect\nthe control plane attacks like rushing, wormhole, black hole, grey hole,\nnetwork partitioning, and routing loop attacks. On the other hand, neigh-\nbor monitoring techniques [26,27] are employed to detect the data plane\nattacks.\n3.6\nToward Standardization\nIEEE 802.11i [30] is the defined standard for the MAC layer security of the\nwireless networks. The draft standard for wireless mesh networks, IEEE\n802.11s, has proposed the use of IEEE 802.11i for the MAC layer security in\nwireless mesh networks. Therefore, we dedicate this section to discuss the\nIEEE 802.11i standard. We first explain the security methods used and the\nsecurity services provided in the IEEE 802.11i standard, and later we will\nexpose the vulnerabilities in IEEE 802.11i that render the standard prone\nto security attacks. These attacks include the pre-computation and partial\nmatching, session hijacking, and the man-in-the-middle attacks exploiting\nvulnerabilities in IEEE 802.1X, and DoS attacks exploiting vulnerabilities in\nthe four-way handshake. We also discuss the proposed prevention mech-\nanisms for these attacks briefly.\nIEEE 802.11i provides the security services of data confidentiality, data\nintegrity, authentication, and protection against replay attacks. The stan-\ndard consists of three components: key distribution, mutual authentication,\nand data confidentiality integrity and origin authentication. In the following\nparagraphs, we briefly discuss these components.\nIEEE 802.1X [33] is used for key distribution and authentication, entailing\nthe use of Extensible Authentication Protocol (EAP) [34] and an authenti-\ncation, authorization, and accounting server (AAA server) like RADIUS or\nDIAMETER [35,36]. IEEE 802.1X is a port-based access control protocol\n" }, { "page_number": 144, "text": "Attacks and Security Mechanisms\n■\n133\nwhich operates in client–server architecture. When the router/access point\n(authenticator) detects a new client (supplicant), the port on the authenti-\ncator is enabled and set to the “unauthorized” state for that client. In this\nstate, only 802.1X traffic (EAP messages) is allowed and all other traffic is\nblocked from that client. The authenticator sends out the EAP-Request mes-\nsage to the supplicant, and the supplicant replies with the EAP-Response\nmessage. The authenticator forwards this message to the AAA server. If\nthe server authenticates the client and accepts the request, it generates a\nPairwise Master Key (PMK), which is distributed to the authenticator and\nthe supplicant using EAP messages. After authentication from the server,\nthe authenticator sets the port for the client to the “authorized” state and\nnormal traffic is allowed. Note that the same protocol can be used to au-\nthenticate and distribute keys between two peer routers or two peer clients\nin case of wireless mesh networks.\nEncryption key distribution and authentication using 802.1X is followed\nby mutual authentication of supplicant (client or peer router) and authen-\nticator (router/AP or peer router), which is based on the four-way hand-\nshake. The four-way handshake is initiated when the two nodes intend to\nexchange data. The encryption key distribution makes the shared secret\nkey PMK available to the supplicant as well as the authenticator. However,\nthis key is designed to last the entire session and should be exposed as lit-\ntle as possible. Therefore the four-way handshake is used to establish two\nmore keys called the Pairwise Transient Key (PTK) and Group Temporal\nKey (GTK). PTK is generated by the supplicant by concatenating the PMK,\nAuthenticator nonce (ANonce), Supplicant nonce (SNonce), Authenticator\nMAC address, and Supplicant MAC address. The product is then put through\na cryptographic hash function. GTK is generated by the authenticator and\ntransmitted to the supplicant during the four-way handshake. PTK is used\nto generate a Temporal Key (TK), which is used to encrypt unicast mes-\nsages while the GTK is used to encrypt broadcast and multicast messages.\nThe four-way handshake (shown in Figure 3.10) involves generation and\ndistribution of these keys between supplicant and authenticator, resulting in\nmutual authentication. The first message of the four-way handshake is trans-\nmitted by the authenticator to the supplicant, which consists of ANonce. The\nsupplicant uses this ANonce and readily available fields with itself to gen-\nerate the PTK. The second message of the handshake is transmitted by the\nsupplicant to the authenticator consisting of SNonce and Message Integrity\nCode (MIC), which is encrypted using PTK. The authenticator is then able\nto generate the PTK and GTK. The attached MIC is decrypted using the gen-\nerated PTK. If the MIC is successfully decrypted, then the authenticator and\nthe supplicant have successfully authenticated each other (Mutual Authenti-\ncation). This is because the authenticator’s generated PTK will only match\nthe PTK transmitted by the supplicant if the two share the same PMK. A third\nmessage is transmitted by the authenticator consisting of GTK and MIC.\n" }, { "page_number": 145, "text": "134\n■\nSecurity in Wireless Mesh Networks\nAuthenticator\nSupplicant\nSNonce, MIC\nGTK, MIC\nACK\nANonce\nConstruct PTK\nConstruct GTK\nFigure 3.10\nFour-way handshake.\nThe last message of the four-way handshake is the acknowledgment trans-\nmitted by the supplicant. The two nodes can exchange the data after a\nsuccessful four-way handshake.\nIEEE 802.11i provides two methods for the security services of data\nconfidentiality, data integrity, origin authentication, and protection against\nreplay attacks. The first method, Temporal Key Integrity Protocol (TKIP), is\nthe enhanced version of WEP and has been provided for backward com-\npatibility with the hardware that was designed to use WEP. RC4 encryption\nhas been used as the encryption algorithm; however, the implementation\nof the algorithm is weak, rendering the protocol vulnerable to numerous\nsecurity attacks. We do not discuss this method in detail. Interested readers\nare referred to Section 8.3.2 of the standard [30] for further details of the\nmethod.\nThe second method is the Counter mode (CTR) with CBC-MAC Protocol\n(CCMP). CCMP is based on the Counter mode with CBC-MAC (CCM) [37] of\nthe AES encryption algorithm. CCM combines Counter (CTR) for confiden-\ntiality and the Cipher Block Chaining (CBC) Message Authentication Code\n(MAC) for origin authentication and integrity. As shown in Figure 3.11, CCM\nencryption takes four inputs: the encryption key, Additional Authentication\nData (AAD), a unique Nonce for every frame, and the plaintext. CCM re-\nquires a fresh TK (generated from PTK) for every session which is used\nas the encryption key. AAD is constructed from the MAC header, and con-\nsists of the following fields: Frame Control field FC (certain bits masked),\nAddress A1, Address A2, Address A3, Sequence Control field SC (certain\nbits masked), Address A4 (if present in the MAC header), and quality-of-\nservice Control field QC (if present). CCMP uses the A2 and the priority\nfields of the MAC header along with a 48-bit packet number (PN) to gener-\nate the unique nonce value for each frame protected by a given TK. PN is\n" }, { "page_number": 146, "text": "Attacks and Security Mechanisms\n■\n135\nIncrement PN\nConstruct CCMP header\n||\nConstruct AAD\nCCM\nencryption\nConstruct\nnonce\nEncrypted\ndata, MIC\nData\nA2, priority\nMAC header\nKeyID\nPN\nTK\nPlain text\nMPDU\nFigure 3.11\nCCMP encryption process and encrypted frame generation [30].\nincremented for each MPDU, resulting in a fresh value of nonce for each\nMPDU. The output of the encryption is the cipher text and the MIC. The\nframe to be transmitted is constructed by concatenating the MPDU header,\nCCMP header, cipher text, and MIC. CCM encryption is explained in\nRFC 3610.\n3.6.1\nVulnerabilities in IEEE 802.11i and Security Attacks\nThe IEEE 802.11i standard successfully provides a number of security ser-\nvices; however, a number of security vulnerabilities have been identified in\nrecent years. We discuss these vulnerabilities, the attacks exploiting these\nvulnerabilities, and the available prevention mechanisms in this sub-section.\n3.6.1.1\nIEEE 802.1X Vulnerabilities\nIEEE 802.1X [33] is used by IEEE 802.11i standard for key distribution and\nauthentication. Three entities, Authenticator, Supplicant, and the Authen-\ntication server, participate in the process. The basic assumption underly-\ning the protocol is that the authenticator is always trusted. Therefore, the\nsupplicant does not verify the messages received from the authenticator\nand unconditionally responds to these messages. However, in practice the\nadversary can also act as authenticator, which renders the protocol vul-\nnerable to session hijacking and man-in-the-middle attacks as exposed in\n[45]. Figure 3.12 shows how an adversary can exploit the above-mentioned\nvulnerability to launch a session hijacking attack. The adversary waits un-\ntil the authenticator and the supplicant complete the authentication pro-\ncess and the authenticator sends the EAP success message to the suppli-\ncant. Following this, the adversary sends a disassociate message to the\n" }, { "page_number": 147, "text": "136\n■\nSecurity in Wireless Mesh Networks\nAuthenticator\nSupplicant\nAdversary\nEAP response\nEAP request\nEAP success\nSupplicant authenticated\n802.11 MAC disassociate\nNetwork traffic\nGains network\nconnectivity\nAuthenticator’s\nMAC address\nspoofed\nFigure 3.12\nSession hijacking attack on 802.1X authentication mechanism.\nsupplicant with the spoofed IP of the authenticator. The supplicant as-\nsumes its session has been terminated by the authenticator as the message\nis not verified for integrity. The adversary gains access to the network by\nspoofing the MAC address of the supplicant and proceeds with a mutual\nauthentication procedure using the four-way handshake.\nFigure 3.13 shows a man-in-the-middle attack launched by the adver-\nsary exploiting the vulnerability in IEEE 802.1X. After the initial exchange\nof EAP request and response messages between the supplicant and the au-\nthenticator, the adversary sends an EAP success message to the supplicant\nusing its own MAC address. Because the IEEE 802.1X protocol suggests\nunconditional transition upon receiving the EAP success message by the\nsupplicant, the supplicant assumes it is authenticated by the authenticator\nAuthenticator\nSupplicant authenticated\nAdversary\nSupplicant\nEAP response\nEAP request\nEAP success\nSupplicant\nstate moves\nahead\nNetwork traffic\nEAP success\nNo action\nFigure 3.13\nMan-in-the-middle attack on 802.1X authentication mechanism.\n" }, { "page_number": 148, "text": "Attacks and Security Mechanisms\n■\n137\nand changes the state. When the authenticator sends the EAP success mes-\nsage, the supplicant has already passed the stage where it was waiting for\nthe success message, and hence no action is taken for this message. The\nsupplicant assumes the adversary as the legitimate authenticator while the\nadversary can easily spoof the MAC address of the supplicant to commu-\nnicate with the actual authenticator. Therefore, the adversary will become\nthe intermediatory between the supplicant and the authenticator. The pro-\nposed solutions to prevent these attacks [45] recommend the authentication\nand integrity check for the EAP messages between the authenticator and\nthe supplicant. The solution also proposes that the peer-to-peer based au-\nthentication model be adopted where the authenticator and the supplicant\nshould be treated as peers and the supplicant should verify the messages\nfrom the authenticator during the process of trust establishment. The peer-\nto-peer model is suitable for WMNs where both the authenticator and the\nsupplicant are WMN routers.\n3.6.1.2\nFour-Way Handshake Vulnerabilities\nFour-way handshake is the mechanism used for the mutual authentication\nof the supplicant and the authenticator in IEEE 802.11i. Vulnerabilities in\nthe four-way handshake have been identified and the DoS attack exploit-\ning these vulnerabilities proposed in [44]. The handshake starts after PMK is\ndistributed to the supplicant and the authenticator. The supplicant waits for\na specific interval of time for message 1 of the handshake from the authenti-\ncator. If the message is not received, the supplicant disassociates itself from\nthe authenticator. Note that this is the only timer used by the supplicant.\nIf message 1 is received by the supplicant, it is then bound to respond to\nevery message from the authenticator and wait for the response until it is re-\nceived. On the other hand, the authenticator will time-out for every message\nif it does not receive the expected response within a specific time interval.\nFurther, the supplicant is de-authenticated if the response is not received\nafter several retries. Also note that both the authenticator and the supplicant\ndrop the message silently if the MIC of the message cannot be verified.\nThis mechanism of four-way handshake is vulnerable to the DoS at-\ntack by the adversary. Consider Figure 3.14 where the authenticator sends\nmessage 1 to the supplicant. Note that message 1 is not encrypted. Suppli-\ncant generates a new SNonce and then generates PTK using the ANonce,\nSNonce, and other relevant fields and responds with the message 2, which\nis encrypted using PTK. At this point, the adversary sends the malicious\nmessage 1 with the spoofed MAC address of the authenticator. The suppli-\ncant is bound to respond to the message. It assumes that the message 2\nthat it sent to the authenticator is lost so the authenticator has retransmit-\nted the message 1. Therefore, it calculates PTK’ (different from PTK and\noverwriting PTK) based on the ANonce sent by the adversary and sends\n" }, { "page_number": 149, "text": "138\n■\nSecurity in Wireless Mesh Networks\nAuthenticator\nAttacker\nSupplicant\nMsg 1: ANonce\nMsg 1: ANonce’\nConstruct PTK’\n(Attacker sends messages with spoofed MAC address of authenticator)\nMsg 3: GTK, MICPTK\nMsg 2: SNonce, MICPTK\nPTK and PTK’ differ\nMIC not verified\nprotocol blocked\nConstruct PTK\nConstruct GTK\n[PTK overwritten]\nFigure 3.14\nDoS attack on four-way handshake. Attacker sends messages with\nspoofed MAC address of authenticator.\nmessage 2 again which is encrypted using PTK’. Meanwhile, the authen-\nticator responds to the first message 2 of the supplicant by sending the\nmessage 3 which is encrypted using PTK. The integrity check performed\nby the supplicant on message 3 fails because the supplicant is now using\nPTK’ while the authenticator encrypted the message using PTK. Conse-\nquently the four-way handshake process is blocked until the authenticator\nde-authenticates the supplicant after several retries, denying the supplicant\nof the service.\nThree solutions have been proposed in [44] to prevent the above attack.\nWe only discuss the most effective solution here. Note that every time\nthe supplicant receives message 1, it generates a new SNonce which is\nconcatenated with ANonce (transmitted by authenticator in message 1) and\nother relevant information to generate new PTK. The proposed solution\nsuggests that the supplicant should store the SNonce created in response\nto the first message 1 that it receives from authenticator. The same SNonce\nshould be used for all subsequent message 1s until the supplicant receives\nmessage 3 from the authenticator. Upon receiving the message 3, supplicant\nshould use the newly transmitted ANonce in message 3 and the stored\nSNonce to generate PTK again, which should be used to decrypt message 3.\nUse of the same SNonce and ANonce will generate the same PTK if other\ninformation remains unchanged. Therefore, the supplicant will be able to\nrespond to the legitimate message 3 even if it receives multiple message\n1s from the adversary. Note that the adversary cannot send a malicious\n" }, { "page_number": 150, "text": "Attacks and Security Mechanisms\n■\n139\nmessage 3 because message 3 is encrypted using PTK, which is dependent\non PMK (only known to the supplicant and the authenticator).\n3.6.1.3\nCCMP Encryption Vulnerabilities\nAlthough CCMP (employed by IEEE 802.11i) uses the CCM encryption, the\nstrength of which is time tested, the protocol is vulnerable to the partial\nmatching and pre-computation attacks. The vulnerabilities of the protocol\nimplementation and the resulting attacks have been exposed in [40]. The\nresearch shows that the address field A2 and the priority field of the MAC\nheader and the PN field in the CCMP header are transmitted as plaintext\nin the headers as well as in the encrypted form as part of the MIC. This\nleads to the partial matching attack and the researchers have shown that\nthe key strength of the 128-bit encryption key used in CCMP decreases. The\ndecrease in the key strength exposes the protocol to pre-computation at-\ntack, resulting in the compromise of data confidentiality and data integrity.\nFurther, the CCM encryption is a two-phase process. During the first phase,\nthe MIC is calculated, and in the second phase, the encryption of the frame\ntakes place. Similarly, the decryption is done in two phases where first the\nmessage integrity is verified from MIC and then the decryption takes place.\nThe two-phase processing of the frame at each wireless link may lead to\nconsiderable delay in case of multi-hop wireless networks like wireless\nmesh networks where the data traverses a number of intermediate wireless\nhops before reaching the wired Internet. The delay introduced by the se-\ncurity services leads to the impracticability of the CCMP protocol for large\nwireless mesh networks consisting of several intermediate hops.\n3.7\nOpen Issues\nA number of security solutions have been discussed aimed at solving dif-\nferent security issues, preventing, detecting, and countering the security at-\ntacks; however, a number of open issues still require considerable\nattention.\n■\nQuite a few intrusion detection systems exist for multi-hop wire-\nless networks; however, very few solutions actually comply with\nthe characteristics of the security solution for WMN (listed in\nSection 3.4). For example, very few solutions lead to the self-healing\nand self-organized WMN, primarily because of the lack of appropri-\nate response mechanism to the detected anomalies and possible\nattacks in the network.\n■\nA number of authentication mechanisms have been proposed for\nmulti-hop wireless networks. However, either the solutions incur\nunacceptable overheads to cater for mobility or the solutions are\n" }, { "page_number": 151, "text": "140\n■\nSecurity in Wireless Mesh Networks\nnon-robust in an effort to accommodate the trade-off between avail-\nable resources and the achievable security level. Neither high mobil-\nity nor the resource limitation is a major design constraint for WMN.\nTherefore, the authentication mechanisms for WMN can be more\nrobust with limited overhead and need to be redefined keeping in\nview the characteristics of WMN.\n■\nAlthough efforts have been made to address the security issues orig-\ninating from colluding malicious nodes that can launch the attacks\nlike wormhole and black hole, no solution has successfully ad-\ndressed the issue of colluding malicious nodes. The malicious and\nmisbehaving nodes pose serious threats to WMN, specifically if the\nnetwork has to be self-healing and self-organized.\n■\nNo security mechanism has so far been proposed to address the\nsecurity vulnerabilities in the joint channel assignment and routing\nalgorithms for multi-radio multi-channel WMN. These algorithms are\ncrucial for the performance of multi-radio multi-channel WMN and a\nsecurity loophole in these algorithms can lead to severely degraded\nperformance and, in some cases, the complete DoS.\n■\nIEEE 802.11i, the standard for security in wireless networks, needs\nto address the issues identified in Section 3.6 before it can be inte-\ngrated into IEEE 802.11s (draft standard for WMN) as the security\ncomponent.\n3.8\nConclusion\nIn this chapter, we considered the security issues in wireless mesh net-\nworks that render these networks vulnerable to security attacks. Different\nsecurity attacks on the MAC layer and network layer of wireless mesh net-\nworks have been considered in detail. Security mechanisms used to detect,\nprevent, and counteract these attacks have been discussed briefly. The in-\ntrusion prevention and detection mechanisms used in various multi-hop\nwireless networks and applicable to wireless mesh networks have been\nconsidered. The IEEE 802.11i standard for security in wireless networks\nhas been discussed in detail along with a note on the vulnerabilities ren-\ndering the protocol impractical for use in wireless mesh networks.\nReferences\n[1]\nAshish Raniwala, Kartik Gopalan, and Tzi-cker Chiueh. Centralized chan-\nnel assignment and routing algorithms for multi-channel wireless mesh\nnetworks. In ACM SIGMOBILE Mobile Computing and Communications\nReview (MC2R), April 2004.\n" }, { "page_number": 152, "text": "Attacks and Security Mechanisms\n■\n141\n[2]\nAshish Raniwala and Tzi-cker Chiueh. Architecture and algorithms for an\nIEEE 802.11-based multi-channel wireless mesh network. In Proceedings of\nIEEE InfoCom, March 2005.\n[3]\nMurali Kodialam and Thyaga Nandagopal. Characterizing the capacity\nregion in multi-radio multi-channel wireless mesh networks. In Proceedings\nof Mobile Computing and Networking, August 2005.\n[4]\nMansoor Alicherry, Randeep Bhatia, and Li (Erran) Li. Joint Channel\nAssignment and Routing for Throughput Optimization in Multi-radio Wire-\nless Mesh Networks. In Proceedings of Mobile Computing and Networking,\nAugust 2005.\n[5]\nKrishna N. Ramachandran, Elizabeth M. Belding-Royer, Kevin C. Almeroth,\nand Milind M. Buddhikot. Interference-aware channel assignment in multi-\nradio wireless mesh networks. In Proceedings of IEEE Infocom 2006, April\n2006.\n[6]\nP. Papadimitratos and Z. Haas. Secure routing for mobile ad hoc networks.\nIn SCS Communication Networks and Distributed Systems Modeling and\nSimulation Conference (CNDS 2002), January 2002.\n[7]\nManel Guerrero Zapata and N. Asokan. Securing ad hoc routing protocols.\nIn Proceedings of the ACM Workshop on Wireless Security (WiSe 2002),\nSeptember 2002.\n[8]\nKimaya Sanzgiri, Bridget Dahill, Brian Neil Levine, Clay Shields, and\nElizabeth Belding-Royer. A secure routing protocol for ad hoc networks. In\nProceedings of the 10th IEEE International Conference on Network Protocols\n(ICNP ’02), November 2002.\n[9]\nYih-Chun Hu, Adrian Perrig, and David B. Johnson. Ariadne: A secure on-\ndemand routing protocol for ad hoc networks. In Proceedings of the 8th\nAnnual International Conference on Mobile Computing and Networking\n(MobiCom 2002), pp. 12–23, September 2002.\n[10]\nHuaizhi Li and Mukesh Singhal. A secure routing protocol for wireless ad\nhoc networks. In Proceedings of the 39th Hawaii International Conference\non System Sciences, January 2006.\n[11]\nYih-Chun Hu, Adrian Perrig, and David B. Johnson. Rushing attacks and\ndefense in wireless ad hoc network routing protocols. In Proceedings of\nthe 2003 ACM Workshop on Wireless Security (WiSe 2003), in conjunction\nwith MobiCom, pp. 30–40, September 2003.\n[12]\nYih-Chun Hu, Adrian Perrig, and David B. Johnson. Packet leashes: A de-\nfense against wormhole attacks in wireless ad hoc networks. In Proceedings\nof IEEE INFOCOM 2003, April 2003.\n[13]\nSanjay Ramaswamy, Huirong Fu, Manohar Sreekantaradhya, John Dixon,\nand Kendall E. Nygard. Prevention of cooperative black hole attacks in\nwireless ad hoc networks. International Conference on Wireless Networks,\npp. 570–575, June 2003.\n[14]\nW. Xu, W. Trappe, Y. Zhang, and T. Wood. The feasibility of launching and\ndetecting jamming attacks in wireless networks. In Proceedings of ACM\nMOBIHOC, 2005.\n[15]\nJ. Pollastre, J. Hill, and D. Culler. Versatile low power media access for\nwireless sensor networks. In Proceedings of ACM Sensys, 2004.\n" }, { "page_number": 153, "text": "142\n■\nSecurity in Wireless Mesh Networks\n[16]\nY. Law, L. Hoesel, J. Doumen, P. Hartel, and P. Havinga. Energy-efficient\nlink-layer jamming attacks against wireless sensor network MAC protocols.\nIn Proceedings of the 3rd ACM Workshop on Security of Ad Hoc and Sensor\nNetworks (SASN 2005).\n[17]\nT. Brown, J. James, and A. Sethi. Jamming and sensing of encrypted wireless\nad hoc networks. In Proceedings of ACM MOBIHOC, May 2006.\n[18]\nArunesh Mishra and William A. Arbaugh. An Initial Security Analysis of the\nIEEE 802.1X Standard, Technical report, University of Maryland, February\n2002.\n[19]\nM. Al-Shurman, S. Yoo, and S. Park. Black hole attack in mobile ad hoc net-\nworks. In Proceedings of the 42nd Annual Southeast Regional Conference,\nHuntsville, Alabama, April 2004.\n[20]\nJ. Newsome, E. Shi, D. Song, and A. Perrig. The Sybil attack in sen-\nsor networks: Analysis and defenses, 3rd International Symposium on\nInformation Processing in Sensor Networks, IPSN 2004, pp. 259–268,\nApril 2004.\n[21]\nAnjum Naveed and Salil S. Kanhere. Security vulnerabilities in channel\nassignment of multi-radio multi-channel wireless mesh networks. In\nProceedings of IEEE GLOBECOM, November 2006.\n[22]\nS. Zhong, L.E. Li, Y.G. Liu, and Y.R. Yang. On designing incentive-\ncompatible routing and forwarding protocols in wireless ad-hoc networks:\nAn integrated approach using game theoretical and cryptographic tech-\nniques. In Proceedings of IEEE MOBICOM, pp. 117–131, August 2005.\n[23]\nN.B. Salem, L. Buttyan, J.-P. Hubaux, and M. Jakobsson, A charging and\nrewarding scheme for packet forwarding in multi-hop cellular networks.\nIn Proceedings of IEEE MobiHoc, p. 1324, June 2003.\n[24]\nY. Liu, H. Man, and C. Comaniciu. A game theoretic approach to efficient\nmixed strategies for intrusion detection. In IEEE International Conference\non Communications (ICC), 2006.\n[25]\nAna Paula R. da Silva, Marcelo H.T. Martins, Bruno P.S. Rocha, Antonio\nA.F. Loureiro, Linnyer B. Ruiz, and Hao Chi Wong. Decentralized intrusion\ndetection in wireless sensor networks. In Proceedings of the 1st ACM In-\nternational Workshop on Quality of Service and Security in Wireless and\nMobile Networks (Q2SWinet 2005), pp. 16–23, October 2005.\n[26]\nYi-an Huang, Wei Fan, Wenke Lee, and Philip S. Yu. Cross-feature analysis\nfor detecting ad-hoc routing anomalies. Proceedings 23rd International\nConference on Distributed Computing Systems, pp. 478–487, May 2003.\n[27]\nYi-an Huang and Wenke Lee. A cooperative intrusion detection system for\nad hoc networks. Proceedings of the 1st ACM Workshop on Security of Ad\nHoc and Sensor Networks, pp. 135–147, October 2003.\n[28]\nHao Yang, J. Shu, Xiaoqiao Meng, and Songwu Lu. SCAN: Self-organized\nnetwork-layer security in mobile ad hoc networks. IEEE Journal on Selected\nAreas in Communications, Volume 24, Issue 2, pp. 261–273, February 2006.\n[29]\nSecurity Architecture for Open Systems Interconnection for CCITT Appli-\ncations, ITU-T Recommendation X.800, March 1991.\n" }, { "page_number": 154, "text": "Attacks and Security Mechanisms\n■\n143\n[30]\nIEEE Std. 802.11i-2004, Wireless Medium Access Control (MAC) and Phys-\nical Layer (PHY) Specifications: Medium Access Control (MAC) Security\nEnhancements, July 2004, http://standards.ieee.org/getieee802/dwnload/\n802.11i-2004.pdf.\n[31]\nHamdy S. Soliman and Mohammed Omari. Application of synchronous\ndynamic encryption system in mobile wireless domains. In Proceedings of\nthe 1st ACM International Workshop on Quality of Service and Security in\nWireless and Mobile Networks (Q2SWinet ’05), pp. 24–30, October 2005.\n[32]\nKui Ren, Wenjing Lou, and Yanchao Zhang. LEDS: Providing location-aware\nend-to-end data security in wireless sensor networks. In Proceedings of IEEE\nInternational Conference on Computer Communication (INFOCOM ’06),\nApril 2006.\n[33]\nIEEE Std. 802.1X-2004, IEEE Standard for Local and Metropolitan Area\nNetworks — Port-Based Network Access Control, June 2001. http://standards.\nieee.org/getieee802/download/802.1X-2004.pdf.\n[34]\nB. Aboba, L. Blunk, J. Vollbrecht, J. Carlson, and H. Levkowetz, Eds.,\nExtensible Authentication Protocol (EAP), RFC 3748, June 2004.\n[35]\nC. Rigney, S. Willens, A. Rubens, and W. Simpson, Remote Authentication\nDial In User Service (RADIUS), RFC 2865, June 2000.\n[36]\nP. Calhoun, J. Loughney, E. Guttman, G. Zorn, and J. Arkko, Diameter Base\nProtocol, RFC 3588, September 2003.\n[37]\nD. Whiting, R. Housley, and N. Ferguson, Counter with CBC-MAC (CCM),\nRFC 3610, September 2003.\n[38]\nS.L. Keoh and E. Lupu. Towards flexible credential verification in mobile\nad-hoc networks. In Proceedings of the 2nd ACM International Workshop\non Principles of Mobile Computing, POMC ’02. Toulouse, France, October\n2002.\n[39]\nJ. Kong, P. Zerfos, H. Luo, S. Lu, and L. Zhang. Providing robust and\nubiquitous security support for MANET. In Proceedings of IEEE ICNP, 2001,\npp. 251–260.\n[40]\nM. Junaid, Muid Mufti, and M. Umar Ilyas. Vulnerabilities of IEEE 802.11i\nwireless LAN CCMP protocol, Transactions on Engineering, Computing and\nTechnology, Volume 11, February 2006.\n[41]\nN. Aboudagga, M.T. Refaei, M. Eltoweissy, L.A. DaSilva, and J. Quisquater.\nAuthentication protocols for ad hoc networks: Taxonomy and research\nissues. In Proceedings of the 1st ACM International Workshop on Quality\nof Service and Security in Wireless and Mobile Networks (Q2SWinet ’05).\nMontreal, Quebec, Canada, October 2005.\n[42]\nD. Hongmei, A. Mukherjee, and D.P. Agrawal. Threshold and identity-based\nkey management and authentication for wireless ad hoc networks, In Pro-\nceedings of International Conference on Information Technology: Coding\nand Computing (ITCC 2004), pp. 107–111, Vol. 1, April 2004.\n[43]\nY.-X. Lim, T.S. Yer, J. Levine, and H.L. Owen. Wireless intrusion detection\nand response. Information assurance workshop 2003. IEEE Systems, Man\nand Cybernetics Society, pp. 68–75, June 2003.\n" }, { "page_number": 155, "text": "144\n■\nSecurity in Wireless Mesh Networks\n[44]\nChanghua He and John C. Mitchell, Analysis of the 802.11i 4-way hand-\nshake, WiSEı04, Philadelphia, October 2004.\n[45]\nArunesh Mishra and A. William Arbaugh, An Initial Security Analysis of\nthe IEEE 802.1X Standard, Technical report CS-TR-4328, Department of\nComputer Science, University of Maryland, February 2002, https://drum.\numd.edu/dspace/handle/1903/1179?mode=full.\n" }, { "page_number": 156, "text": "Chapter 4\nIntrusion Detection in\nWireless Mesh Networks\nThomas M. Chen, Geng-Sheng Kuo, Zheng-Ping Li,\nand Guo-Mei Zhu\nContents\n4.1\nIntroduction ........................................................ 146\n4.2\nIntrusion Detection ................................................. 148\n4.2.1\nGoals of Intrusion Detection .............................. 149\n4.2.2\nHost-Based and Network-Based Monitoring .............. 149\n4.2.3\nMisuse Detection and Anomaly Detection ................ 150\n4.2.4\nIDS Response .............................................. 152\n4.3\nUnique Challenges of Wireless Mesh Networks................... 152\n4.3.1\nWireless Medium........................................... 153\n4.3.2\nDynamic Network Topology .............................. 154\n4.4\nIntrusion Detection for Wireless Mesh Networks ................. 154\n4.4.1\nWATCHERS ................................................. 154\n4.4.2\nCooperative Anomaly Detection .......................... 155\n4.4.3\nWatchdogs and Pathraters ................................. 156\n4.4.4\nTIARA ...................................................... 157\n4.4.5\nMalcounts .................................................. 157\n4.4.6\nCONFIDANT ............................................... 158\n4.4.7\nMobIDS..................................................... 159\n4.4.8\nMobile Agents.............................................. 160\n4.4.9\nAODVSTAT ................................................. 160\n4.4.10\nTrust Model ................................................ 161\n4.4.11\nRESANE .................................................... 162\n145\n" }, { "page_number": 157, "text": "146\n■\nSecurity in Wireless Mesh Networks\n4.4.12\nCritical Nodes .............................................. 162\n4.4.13\nSCAN ....................................................... 163\n4.4.14\nDempster–Shafer ........................................... 164\n4.4.15\nOptimization of Limited Resources........................ 164\n4.5\nOpen Research Issues .............................................. 165\n4.5.1\nLack of Experience with Wireless Mesh Networks ....... 165\n4.5.2\nEvaluation Difficulties ..................................... 165\n4.5.3\nIntrusion Tolerance ........................................ 166\n4.6\nConclusion.......................................................... 166\nReferences................................................................. 167\nWireless mesh networks are potentially vulnerable to a broad variety of\nattacks. Hence security is an important consideration for the practical oper-\nation of wireless mesh networks. Within security, intrusion detection is the\nsecond line of defense in wireless networks as well as wired networks. Un-\nfortunately, wireless mesh networks present additional challenges due to\ntheir decentralized nature, dynamic network topology, and easy access to\nthe radio medium. Due to these unique challenges, intrusion detection\ntechniques cannot be borrowed straightforwardly from wired networks.\nNew distributed intrusion detection schemes must be designed for wireless\nmesh networks. This chapter describes the basics of intrusion detection\nand gives a survey of intrusion detection schemes proposed for wireless\nmesh networks. The schemes share some common concepts, but differ in\nthe details which are compared. This chapter describes the difficulties with\neach scheme and ongoing challenges. Due to the difficult challenges pre-\nsented by the wireless environment, intrusion detection in wireless mesh\nnetworks is still an open research problem.\n4.1\nIntroduction\nThe main goal of networks is to relay data between their users. Usabil-\nity in terms of quality of service, availability, and reliability is a typical\ndesign objective. The value of a network is perceived by the services it\nprovides to its users. Unfortunately, security is often a secondary consider-\nation and somewhat contradictory to usability because it usually imposes\naccess restrictions and usage policies. Consequently, many networks are\ninadequately safeguarded against a variety of attacks. Attackers may use\nthe network to direct attacks at hosts (e.g., to access or control a host), or\nattackers may aim to damage the network itself.\nAttacks are commonplace and readily seen on the Internet today [1]. The\naverage PC user must be aware of good security practices, such as keeping\nup with operating system patches, running anti-virus software, turning on\n" }, { "page_number": 158, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n147\na personal firewall, and avoiding suspicious e-mail attachments. Many of\nthese attacks will eventually cross over to wireless networks as well. For\nexample, many attacks exploit vulnerabilities (weaknesses) in operating\nsystems and applications; these are effective in wired or wireless networks.\nAlso, new types of attacks are evolving constantly.\nTypical examples of attacks against hosts include:\n■\nProbing for vulnerabilities\n■\nExploiting vulnerabilities to gain unauthorized access\n■\nEavesdropping on communications\n■\nTheft or alteration of data\n■\nInstallation of malicious software (e.g., viruses, worms, Trojan horses,\nspyware)\n■\nDenial of service\n■\nSocial engineering\n■\nSession hijacking\nSome common attacks against the network include:\n■\nDenial of service against a router or server\n■\nInterception or modification of packets\n■\nInterference with routing protocols\n■\nUnauthorized tampering of Web, DNS (Domain Name System), or\nother servers.\nWireless networks are more vulnerable than wired networks because\nthe wireless medium is shared and accessible through the air. In a wired\nnetwork, an attacker needs to physically access the network to sniff or inject\ntraffic. In a wireless network, an attacker can listen to or transmit packets\non a radio link at a distance (and possibly not in visible sight). Thus, the\nradio medium makes wireless networks both more attractive as targets and\nharder to defend.\nIn addition, the mobility afforded by wireless networks is great for users\nbut has certain implications for security. First, mobile devices tend to travel\nto different, perhaps unfriendly locations. A mobile device is harder to\nphysically secure than a stationary device in a controlled environment.\nWithout adequate physical protection, mobile devices could be physically\ncompromised. Second, mobile users are more difficult to authenticate. A\nstationary user will always access the network at a known location, so\nauthentication can be based at least in part on location (e.g., a landline\nphone is identified by its location). A mobile user may access the network\nat unpredictable locations at different times.\nA mobile ad hoc network (MANET) without any fixed infrastructure\npresents even more challenges for security. With a fixed infrastructure,\nmobile users could be authenticated with an authentication server that is\n" }, { "page_number": 159, "text": "148\n■\nSecurity in Wireless Mesh Networks\nalways accessible regardless of the user’s location. However, in a MANET\nwith a dynamic network topology, nodes may be disconnected from other\nnodes for periods of time. A centralized authentication server would not\nwork because it may not be always reachable from a mobile user’s location.\nWithout the capability for authentication, impersonation attacks are a\nmajor concern in wireless mesh networks. By impersonation, a malicious\nattacker could participate in the dynamic routing protocol and affect the\nchoice of routes. Wireless mesh networks depend on the cooperation of all\nnodes to relay packets across the network, so the integrity of the routing\nprotocol is paramount. The effect of an attack on routing could be degrada-\ntion of network performance, denial of service, or funneling traffic through\nmalicious nodes. Not surprisingly, a great deal of attention has been given\nto secure routing protocols [2–8].\nA unique type of attack called a wormhole has been identified [9]. In\nphysics, a wormhole is theoretically a direct shortcut between two distant\npoints in the space–time continuum. The idea of a wormhole attack is that\npackets at one location in the network could be tunneled and quickly re-\nplayed at another location. A wormhole could be exploited in various ways.\nFor example, it has been hypothesized that routing update packets could go\nthrough a wormhole and cause a routing protocol to avoid certain routes [9].\nDespite the popular stereotype of a misfit teenage “hacker,” there is\nno “typical” attacker or single motive for malicious attacks. An attacker\ncould be almost anyone — a youth looking for fame, a criminal looking for\nprofit, an acquaintance seeking revenge, a competitor attempting industrial\nespionage, or a hostile foreign military agency. One of the difficulties in\nnetwork security (both wired and wireless) is the wide range of types of\nattackers and attack methods.\nOn the defense side, network security consists of a variety of protec-\ntive measures usually deployed in a defense-in-depth strategy. Defense-in-\ndepth refers to multiple lines of defense, such as encryption, firewalls, in-\ntrusion detection systems, access controls, anti-virus and anti-spyware pro-\ngrams, combined together to increase the barriers and costs for attackers.\nThe common belief is that a single perfect defense is not feasible. Instead,\nan effective deterrent can be constructed from multiple lines of defense,\neven though each individual element of defense is imperfect. Intrusion\ndetection is one of the most fundamental elements in a defense-in-depth\nstrategy.\n4.2\nIntrusion Detection\nIntrusion detection can be viewed as a passive defense, similar to a burglar\nalarm in a building. Unlike firewalls or access controls, intrusion detection\nsystems (IDSs) are not intended to deter or prevent attacks. Instead, their\n" }, { "page_number": 160, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n149\npurpose is to alert system administrators about possible attacks, ideally\nin time to stop the attack or mitigate the damage [10]. Because wireless\nnetworks are easier to attack than wired networks, intrusion detection is\nmore critical in wireless networks as a second line of defense.\n4.2.1\nGoals of Intrusion Detection\nIntrusion detection is generally a difficult problem [11]. An IDS attempts\nto differentiate abnormal activities from normal ones, and identify truly\nmalicious activities (attacks) from the abnormal but non-malicious activities.\nUnfortunately, normal activities have a wide range, and attacks may appear\nsimilar to normal activities. For example, a ping is a common utility to\ndiscover if a host is operating and online, but a ping can also be used for\nattack reconnaissance to learn information about potential targets. Even if\nunusual activities can be distinguished from normal activities, an unusual\nactivity may not be truly malicious in intent.\nThe accuracy of intrusion detection is generally measured in terms of\nfalse positives (false alarms) and false negatives (attacks not detected). IDSs\nattempt to minimize both false positives and false negatives. However, this\ngoal is complicated by the likelihood that a skillful attacker will try to evade\ndetection. Thus, detection must be done in adversarial conditions where\nthe attacker may be intelligent and resourceful.\nIDSs also attempt to raise alarms while an attack is in progress, so that\nthe attack can be stopped to minimize damage or the attacker can be\nidentified “in the act.” This goal is difficult considering that attacks may\nconsist of a sequence of inconspicuous steps; many events (e.g., packets)\nmust be analyzed in real-time, and an attack may be new and different\nfrom past experiences.\n4.2.2\nHost-Based and Network-Based Monitoring\nAn IDS essentially consists of three functions, as shown in Figure 4.1 [12].\nFirst, an IDS must collect data by monitoring some type of events. IDSs can\nbe classified into two types depending on the monitored events: host-based\nor network-based IDSs. Host-based IDSs are installed on hosts and monitor\ntheir internal events, usually at the operating system level. These internal\nevents are the type recorded in the host’s audit trails and system logs.\nIn contrast, network-based IDSs monitor packets in the network [13–16].\nThis is usually done by setting the network interface on a host to promis-\ncuous mode (so all network traffic is captured, regardless of packet add-\nresses). Alternatively, there are also specialized protocol analyzers designed\nto capture and decode packets at full link speed.\nA popular network-based IDS is the open-source Snort [17]. In its sim-\nplest mode, Snort can function as a packet sniffer to view packets traversing\n" }, { "page_number": 161, "text": "150\n■\nSecurity in Wireless Mesh Networks\nResponse\nAnalysis engine\nMonitor events\nFigure 4.1\nFunctions of IDS.\na transmission link. In packet logging mode, Snort is able to sniff and dump\ncomplete packets into a log for later analysis. Alternatively, Snort config-\nured with a ruleset can function as a real-time IDS. A Snort ruleset is a\nfile of attack signatures that are matched to captured packets. A match to\na signature means that an attack is recognized. It is essentially a pattern\nmatching technique. Other popular network-based IDSs are Tcpdump and\nEthereal®.\nThe second functional part of an IDS is an analysis engine that processes\nthe collected data. It is programmed with certain intelligence to detect un-\nusual or malicious signs in the data (elaborated below).\nThe third functional part of an IDS is a response, which is typically\nan alert to system administrators. A system administrator is responsible for\nfollow-up investigation of an event after receiving an alert.\n4.2.3\nMisuse Detection and Anomaly Detection\nAs mentioned above, the second functional part of an IDS is an analysis\nengine. Analysis can be done manually by a security expert, but automated\nanalysis is much faster and efficient. The problem with automated analysis\nis programming the analysis engine with a level of intelligence equivalent\nto the knowledge and experience of a security expert.\nCurrently, there are two basic approaches to analysis: misuse detection\nand anomaly detection. Misuse detection is also called signature-based de-\ntection because the idea is to represent every attack by a signature (pattern\nor rule of behavior). Rules can be divided into single part (atomic) sig-\nnatures or multi-part (composite) signatures. It is essentially a problem of\nmatching the observed traffic to signatures. If a matching signature is found,\nthat attack is detected.\nA common implementation of misuse detection is an expert system.\nAn expert system consists of a knowledge base containing descriptions of\nattack behavior based on past experiences and rules that allow matching\nof packets against the knowledge base. These rules are often structured as\n“if-then-else” statements.\n" }, { "page_number": 162, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n151\nAn advantage of misuse detection is its accuracy. If a signature matches,\nthat signature identifies the specific attack. Knowledge of the specific type\nof attack means that an appropriate response can be determined immedi-\nately. For its accuracy, misuse detection is widely preferred in commercial\nsystems.\nThere are two major drawbacks to misuse detection. First, new sig-\nnatures must be developed whenever a new attack is discovered. Cur-\nrently, new attacks are evolving constantly. This means that signatures for\nIDSs must be updated frequently. Second, an attack is recognized only if a\nmatching signature exists. A signature will not exist for new attacks that are\nsignificantly different from known attacks. Thus, misuse detection could\nhave a high rate of false negatives (missed attacks).\nAnomaly detection, sometimes called behavior-based detection, is the\nopposite of misuse detection, as shown in Figure 4.2. Although they are op-\nposite approaches, they can be used together to realize the advantages of\nboth approaches. Misuse detection tries to characterize attacks, and every-\nthing else is assumed to be normal. In contrast, anomaly detection tries to\ncharacterize normal behavior, and everything else is assumed to be anoma-\nlous (although not necessarily malicious). The underlying premise is that\nmalicious activities will deviate significantly from normal behavior.\nThe characterization of normal behavior is called a normal profile. A\nnormal profile is usually constructed by statistical analysis of training data.\nTraining data is typically obtained from observations of past normal behav-\nior. Thus, a normal profile is a statistical picture of past normal behavior. Sig-\nnificant deviations from the normal behavior are deemed to be anomalous.\nAn underlying assumption is that normal behavior will remain the same\nor at least not change quickly. Because real behavior does change over\ntime, practical anomaly detection systems should adapt the normal profile\nto track normal behavior changes. This means practical systems should\nhave a capability for automated learning.\nA major advantage of anomaly detection is the potential to detect new\nattacks without prior experience. That is, a signature for a new attack is not\nrequired; a new attack will be recognized if it significantly deviates from\nnormal behavior.\nAnomaly detection\nMisuse detection\nNormal\nAttack\nsignatures\nAnomalous\nNormal\nprofile\nFigure 4.2\nMisuse detection and anomaly detection.\n" }, { "page_number": 163, "text": "152\n■\nSecurity in Wireless Mesh Networks\nThere are at least three drawbacks to anomaly detection. First, it has\nproven to be extremely difficult in practice to accurately characterize normal\nbehavior because normal activities can have wide deviations. The choices\nof statistical metrics for an accurate profile is still an open research problem.\nSecond, anomalous behavior is not necessarily malicious. In fact, a small\nfraction of anomalous activities may turn out to be an attack. Thus, anomaly\ndetection often shows a high rate of false negatives. These false alarms must\nbe investigated by system administrators, which is time consuming. Third, a\ndetected anomaly does not identify a specific attack, unlike a signature. The\nlack of specific information means that system administrators must perform\na follow-up investigation to determine whether an actual attack is occurring.\n4.2.4\nIDS Response\nAs mentioned above, the third functional component of an IDS is the\nresponse. Detection of an intrusion must lead to some type of output.\nGenerally, responses can be passive or active. An example of a passive\nresponse is to log the intrusion information and raise an alert to system\nadministrators. The IDS does not attempt to impede or stop the intrusion.\nAn IDS response is usually passive because it is widely believed that hu-\nman judgment (by a trained administrator) is required to formulate the most\nappropriate course of action. Also, a system administrator often needs to\nperform further investigation to identify the root cause of an IDS alert.\nActive responses attempt to limit the damage of an attack or stop an\nattack in progress. Damage can be mitigated by protecting the valuable\nassets or the specific target of the attack. Another active response could be\nto track the source of the attack, which might be difficult if the attack is\nbeing carried out through intermediaries. For example, a distributed denial-\nof-service (DDoS) attack is essentially a flooding attack. The flooding traffic\nusually comes from innocent computers that were surreptitiously compro-\nmised by the real attacker. A DDoS attack might be traced to the flooding\ncomputers, but it is difficult to trace the attack further back.\nThere is a risk in tying active responses to intrusion detection, an\napproach called intrusion prevention. In the event of false positives, nor-\nmal traffic is mistakenly identified as malicious. This would trigger an active\nresponse which could cause damage to an innocent user.\n4.3\nUnique Challenges of Wireless Mesh Networks\nIntrusion detection is a common practice in wired networks. Deployment of\nIDS is well understood and relatively straightforward because the network\nenvironment is static. Traffic is relayed by stationary routers. Normally,\nthere are natural points of traffic concentration which are logical candidates\n" }, { "page_number": 164, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n153\nfor monitoring. For example, private organizations usually connect to the\npublic Internet through a gateway and firewall. All incoming and outgoing\ntraffic go through this point. An IDS just outside of the firewall will be able\nto see attacks coming from the untrusted Internet. This is informative for\nunderstanding the external threats that the firewall is intended to block.\nAnother IDS inside the firewall would monitor the traffic in the private\nnetwork. If the firewall is effective, no attacks from the outside should be\ndetected. Obviously any detected intrusion means either an insider attack\nor an external attack penetrated the firewall.\nIn comparison with wired networks, wireless mesh networks present\ndifficulties for intrusion detection. As a review, wireless mesh networks\nhave sprung from MANETs. MANETs have no fixed infrastructure. All nodes\nare mobile and the network topology is dynamic. Nodes are simultaneously\nuser devices and routers. The requirements for MANETs have been driven\nlargely by military or specialized civilian applications [18].\nWireless mesh networks relax the requirement of no fixed infrastruc-\nture, and can have a mix of fixed and mobile nodes interconnected by\nwireless links. As in MANETs, mesh nodes can be simultaneously user de-\nvices and routers. Nodes might also be fixed wireless routers, e.g., IEEE\n802.11 access points or 802.16 subscriber stations [19]. These nodes could\nconstitute a backbone infrastructure [20,21]. A principal characteristic is\nmulti-hop routing, where packets traverse the network by opportunistic\nrelaying from node to node. Multi-hop routes through a wireless mesh\nnetwork are computed by MANET dynamic routing protocols.\n4.3.1\nWireless Medium\nThe wireless medium is one of the major factors affecting intrusion detec-\ntion. In wired networks, traffic is forced to travel along links, and there\nare natural points of traffic concentration which are convenient locations\nfor intrusion detection. This is not as valid in a wireless mesh network,\nparticularly if it is entirely ad hoc. However, there might be a backbone of\nfixed wireless routers. In that case, the traffic through access points should\nbe monitored. In practice, this is difficult because access points typically\ndo not have “SPAN ports” that mirror the traffic.\nMonitoring traffic by promiscuously eavesdropping on the radio medium\nis not ideal. Nodes in a wireless mesh network may have relatively short\nradio ranges (just long enough to reach the next node), so sensors are able\nto see only limited amounts of traffic. Multiple sensors need to be deployed\naround the entire network for a comprehensive view of traffic.\nAnother difficulty presented by the wireless medium is the mobility\nafforded to nodes. As mentioned earlier, mobile devices might travel to\nhostile environments. A mobile device without adequate protection could\nbe physically compromised. Therefore, nodes in a wireless mesh network\n" }, { "page_number": 165, "text": "154\n■\nSecurity in Wireless Mesh Networks\nare more vulnerable to compromise and cannot be entirely trusted even if\ntheir identity is authenticated.\n4.3.2\nDynamic Network Topology\nAgain, the dynamic topology of wireless mesh networks means that there\nare no natural fixed points of traffic concentration which would be good\nchoices for monitoring.\nA possile approach is to run an IDS on certain hosts to monitor their\nlocal neighborhoods. However, a node cannot be expected to monitor the\nsame area for a long time due to its mobility. A node may be unable to\nobtain a large sample of data for accurate intrusion detection.\n4.4\nIntrusion Detection for Wireless Mesh Networks\nNot surprisingly, most of the research in intrusion detection pertains to\nMANETs because wireless mesh networks are a relatively recent develop-\nment. However, virtually all of the intrusion detection schemes for MANETs\nare relevant to wireless mesh networks.\nThis section reviews intrusion detection schemes in chronological order\nto show the evolution of ideas over time; also, see the survey [22].\n4.4.1\nWATCHERS\nNodes in a wireless mesh network relay data in a cooperative way simi-\nlar to the way that Internet routers relay IP packets. Therefore, intrusion\ndetection in the Internet environment has direct relevance to intrusion de-\ntection in wireless mesh and ad hoc networks. One of the earliest intrusion\ndetection schemes proposed for the Internet environment was WATCHERS\n(Watching for Anomalies in Transit Conservation: a Heuristic for Ensuring\nRouter Security) [23]. Although WATCHERS was not specifically intended\nfor ad hoc networks, all nodes in ad hoc networks function as routers,\nso the WATCHERS approach is easily applicable. Later intrusion detection\nschemes for ad hoc networks have followed similar ideas from WATCHERS.\nWATCHERS assumes a wired mesh network of routers where individual\nrouters may be compromised by an attacker or malfunctioning due to a\nfault or misconfiguration. In either case, it is assumed that an intrusion\nor malfunction will be manifested in the router’s misbehavior (selectively\ndropping or misrouting packets) that can be observed by other routers.\nOne of the important ideas of WATCHERS is a totally distributed intru-\nsion detection scheme running concurrently and independently in every\nrouter. Each router checks incoming packets to detect any routing anoma-\nlies. Also, each router keeps track of the amount of data going through\n" }, { "page_number": 166, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n155\nneighboring routers. The objective is to detect misbehaving routers in a\ndistributed way.\nA link-state routing protocol is assumed. This assumption is necessary\nso that each router is aware of other routers and the overall network topol-\nogy. Each router counts any packets that are misrouted by neighboring\nrouters, based on knowledge of their neighbors’ routing tables from the\nlink-state routing protocol. Each router also keeps count of the amount of\ndata received and transmitted on all interfaces.\nRouters periodically share their respective data by a flooding proto-\ncol, and then start a diagnostic phase. Flooding is necessary to overcome\nany malicious nodes that might try to interfere in the information sharing\nby blocking packets. In the diagnostic phase, the counts collected from\nall routers are compared to determine if any routers (1) have misrouted\ntoo many packets, (2) have not participated correctly in the WATCHERS\nscheme, (3) broadcasted counts that have discrepancies with the counts\nfrom their neighbors, and (4) have appeared to drop more packets than a\ngiven threshold. If a router is found to exhibit any of these misbehaviors,\nit is deemed to be a bad router (but it is impossible to determine if the\ncause is an intrusion or malfunction, based solely on the router’s external\nbehavior). In response to any routers deemed to be misbehaving, routing\ntables at good routers are changed to avoid forwarding packets through\nthose misbehaving routers.\nThe counts are compared to thresholds. In an ideal world, the thresholds\nwould be zero, but in practice, the thresholds should be chosen to be more\nthan zero. For example, even good routers may drop a significant number\nof packets if the router is congested. Therefore, the threshold for number\nof dropped packets could be high. The choice of proper thresholds can\nbe difficult. If the thresholds are too high, misbehaving routers could be\nundetectable. On the other hand, if thresholds are too low, the rate of false\nalarms could be significant.\nThere are costs involved in the WATCHERS scheme. Each router must\nuse memory to keep counts and a routing table for each neighboring router.\nAlso, all routers are involved in a flooding protocol to share information\nbefore each diagnostic phase. Moreover, the scheme requires certain con-\nditions to work: (1) each good or bad router must be directly connected\nto at least one good router, (2) each good router must be able to send a\npacket to each other good router through a path of good routers, and (3)\nthe majority of routers must be good.\n4.4.2\nCooperative Anomaly Detection\nOne of the earliest intrusion detection schemes for ad hoc networks was\nproposed by Zhang and Lee [24,25]. One of the basic ideas is distributed\nmonitoring and cooperation among all nodes, similar to the basic idea in\n" }, { "page_number": 167, "text": "156\n■\nSecurity in Wireless Mesh Networks\nWATCHERS. Each node independently observes its neighborhood (within\nits radio range) looking for signs of intrusion. Each node runs an IDS agent\nwhich keeps track of internal activities on that node and packet communi-\ncations within its local neighborhood.\nA second idea in the scheme is to rely mainly on anomaly detection\nbecause of perceived difficulties with misuse detection. Misuse detection\nis limited to the set of known attacks with existing signatures. Also, signa-\ntures must be constantly updated, which would be a difficult process in a\nwireless ad hoc network. Because anamaly detection does not require the\ndistribution of signatures, it is easier to implement in independent nodes.\nEach node develops a normal profile during a training period, and looks\nfor significant deviations from the normal profile to detect anomalies.\nA third idea in the scheme is cooperation among nodes to cover a\nbroader area. If a node has strong evidence of an anomaly, it can raise\nan alert itself. However, if a node has weak or inconclusive evidence of\nan anomaly, it can request a global investigation. The requesting node\nshares its data about the suspected intrusion with its neighboring nodes.\nThe neighboring nodes share their relevant data, and each participating\nnode follows a consensus algorithm to determine whether to raise an alarm.\nAny node that comes to the conclusion that an intrusion exists can raise an\nalarm.\nThe response to an alarm might be recomputation of routing tables\nto avoid compromised nodes, or communication links between nodes are\nforced to re-initialize (re-authenticate each other). The latter would not be\neffective if an attack has compromised a node and captured its authentica-\ntion credentials.\n4.4.3\nWatchdogs and Pathraters\nThe idea of nodes monitoring the packet forwarding behavior of neighbor-\ning nodes was also proposed by Marti et al. [26]. Dynamic source routing\nis assumed. When a packet is ready to be sent, a path to the destination\nis discovered on demand, and the addresses of the nodes along the path\nare encapsulated in the packet header. Two new ideas are introduced:\nwatchdogs and pathraters.\nA watchdog is a process running on a node to monitor the behavior of\nneighboring nodes. After a node forwards a packet, the watchdog monitors\nthe next node to see that the packet is forwarded again. With source routing\nassumed, the watchdog has knowledge of the proper route for a tracked\npacket. If a neighboring node is observed to drop more packets than a\ngiven threshold, that node is deemed to be misbehaving.\nThe pathrater works to avoid routing packets through misbehaving\nnodes. Each node maintains a rating for every other node in the range\nfrom 0 to 1. It calculates a path metric by averaging the node ratings in the\n" }, { "page_number": 168, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n157\npath. Node ratings are initialized to a neutral value of 0.5. Actively used\npaths are incremented periodically, but nodes suspected of misbehaving\nwill have their rating lowered severely.\nBecause the watchdog is a rather simple monitoring process, several\nlimitations were noted. First, the scheme is limited to source routing be-\ncause the watchdog needs knowledge of the proper route for each packet.\nSecond, it is vulnerable to interference by a malicious node falsely reporting\nother nodes as misbehaving. Third, multiple misbehaving nodes could col-\nlectively interfere with the watchdog process. Lastly, a misbehaving node\ncould escape detection by dropping packets just below the threshold level.\n4.4.4\nTIARA\nTIARA (Techniques for Intrusion-resistant Ad hoc Routing Algorithms) was\nactually a set of mechanisms to ensure an ad hoc network could continue\nto operate under hostile adversarial conditions, rather than an intrusion\ndetection scheme [27]. However, a flow monitoring mechanism in TIARA\nis designed to detect path failures from misbehaving nodes.\nThe basic idea is for source nodes to periodically send special “flow\nstatus” messages to destination nodes. Flow status messages contain infor-\nmation about the number of packets that have been sent from the source\nto destination since the previous flow status message. To prevent interfer-\nence with flow status messages, each message is numbered sequentially\n(to detect loss) and encrypted with a digital signature (for authentication).\nUpon receiving a flow status message, the destination node compares\nthe carried number to the actual number of packets received since the last\nflow status message. A path failure is notified to the source node if (1)\na flow status message has been lost or not received by a specified time\ninterval, (2) the actual number of received packets is less than a threshold\nfraction of the number indicated by the source, or (3) the actual number of\nreceived packets is much more than the number indicated by the source.\nThere are two obvious disadvantages of this scheme for intrusion de-\ntection. First, a path failure does not identify which specific nodes could be\ncompromised. Second, the flow status messages incur a cost in additional\ntraffic that is proportional to the number of source-destination pairs in the\nnetwork.\n4.4.5\nMalcounts\nAnother distributed intrusion detection system proposed by Bhargava and\nAgrawal [28] is essentially an enhancement of Zhang and Lee’s approach. As\nbefore, it is assumed that each node is independently and concurrently\nmonitoring its local neighborhood (nodes within its radio range). AODV\n" }, { "page_number": 169, "text": "158\n■\nSecurity in Wireless Mesh Networks\n(Ad hoc On-demand Distance Vector) routing is assumed. When a packet is\nready to be sent, the source node will flood a request through the network;\na request successfully reaching the destination will be acknowledged back\nto the source.\nThe central idea in the intrusion detection scheme is that each node\nmaintains a “malcount” for neighboring nodes, which is the number of ob-\nserved occurrences of misbehavior. When the malcount for a node exceeds\na given threshold, an alert is sent out to other nodes. The other nodes then\ncheck their malcounts for the suspected node and may support the initial\nalert with secondary alerts. If a suspected node triggers two or more alerts,\nit is deemed to be malicious and a “purge” message is broadcasted. In\nresponse, the suspected node is avoided by the other nodes.\nA problem with the proposed scheme is that it is not clear if malcounts\nare only cumulative, so they can increase but not decrease. The ability to\ndecrease malcounts would be useful for nodes with unusual but not mali-\ncious behavior that might be falsely identified as malicious. Their unusual\nbehavior might cause their malcount to increase, but then a period of good\nbehavior would result in their malcount returning to a normal value. This\ncould avoid false alerts.\nNaturally, this scheme works only if at least two trustworthy nodes are\nobserving a suspected node, and can be defeated by malicious nodes send-\ning out false alerts. Also, the scheme depends on a threshold for malcounts.\nA compromised node could avoid detection by keeping its misbehavior\nunder the threshold.\n4.4.6\nCONFIDANT\nThe CONFIDANT (Cooperation of Nodes: Fairness in Dynamic Ad hoc\nNetworks) scheme was proposed by Buchegger and Le Boudec [29]. Like\nprevious schemes, it is highly distributed with each node monitoring its lo-\ncal neighborhood and cooperatively sharing information with other nodes.\nSource routing is assumed so that nodes have knowledge of the correct\nroute for tracked packets. In each node, the CONFIDANT system includes\nfour components: the monitor, reputation system, trust manager, and path\nmanager.\nSimilar to Zhang and Lee’s approach, the monitor in each node observes\nthe activities of neighboring nodes (within radio range) to look for misbe-\nhavior. With source routing assumed, the monitor has knowledge of the\nnext hop for each packet. When the node forwards a packet to a neighbor,\nit watches the neighbor to see whether the packet is forwarded correctly to\nthe next hop. A copy of the entire packet is also stored temporarily to de-\ntect any suspicious modifications to the forwarded packet. If a misbehavior\nis observed, the reputation system is called.\n" }, { "page_number": 170, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n159\nThe reputation system is similar in concept to Bhargava and Agrawal’s\nmalcount and Marti et al.’s node ratings. The reputation system consists of\na table listing all observed nodes and their reputation ratings. If a node is\nobserved to be misbehaving (deviating from expected routing behavior),\nthe node’s rating is changed by a weighting function, depending on the\nconfidence in the accuracy of the new observation. To reduce the chance\nof false alarms, a node’s rating can be improved after a specified period of\ngood behavior. If a node’s rating falls below a threshold, the path manager\nis called.\nThe path manager has a number of responsibilities. It keeps track of\na security rating for paths, depending on the reputations of nodes in the\npath. Paths containing a malicious node are deleted. If a received packet\nis going on a path containing a malicious node, the packet is ignored and\nthe source is alerted. If a received packet comes from a malicious node,\nthe packet is ignored.\nThe last component, the trust manager, is responsible for receiving\nand sending “alarm” messages. Alarm messages contain information about\nobserved misbehaviors to warn about suspected nodes. Alarm messages\nare sent to other nodes on a “friends” list, although the maintenance of the\nfriends list has not been described. When a node receives an alarm mes-\nsage, the trust manager looks up the source of the message. If the source is\ntrusted, the alarm message is added to a table of alarms. If there is enough\nevidence that a reported node is indeed malicious, the information is passed\nto the reputation manager.\nA number of details in the CONFIDANT scheme remain to be developed.\nFor example, misbehaviors besides incorrect packet forwarding are not\nyet specified. Other missing details are the values for thresholds, time-out\nfor improving reputations, and who qualifies for the friends list. Also, the\nscheme is currently limited to source routing.\n4.4.7\nMobIDS\nMobIDS (Mobile Intrusion Detection System) proposed by Kargl et al. [30] is\ngenerally similar to the previous distributed IDS schemes. Multiple sensors\nin the network keep track of observed instances of cooperative and non-\ncooperative behavior of nodes. Cooperative instances are given positive\nvalues whereas non-cooperative instances are given negative values. All\ninstances observed for a suspected node are combined to calculate a sensor\nrating for that node, where older instances are given less weight. Then all\nsensor ratings for a suspected node are averaged, with a weight reflecting\nthe credibility of each sensor, into a “local rating” for that node.\nThe local ratings are distributed periodically by broadcasting them to\nneighboring nodes within a given range. Each node averages the local\n" }, { "page_number": 171, "text": "160\n■\nSecurity in Wireless Mesh Networks\nratings that it receives (including its own rating) into global ratings for\nother nodes. But global ratings are accepted only when at least a prespec-\nified minimum number of nodes have contributed to the rating. Nodes are\ndeemed to be misbehaving if their ratings drop below a given threshold.\nRoutes are changed to avoid misbehaving nodes, and packets related to\nthose nodes are dropped.\n4.4.8\nMobile Agents\nPuttini et al. [31] propose a distributed IDS scheme that is similar archi-\ntecturally to previous proposals except that mobile agents are used for\ninteractions between nodes (instead of data). Mobile agents are software\nprograms that can autonomously suspend execution at one node, transfer\ntheir code and state to another node, and resume execution at the sec-\nond node. Mobile agents are usually implemented in JavaTM because the\nJava Virtual Machine is widely supported on a broad variety of operating\nsystems.\nEach node independently runs a process called local IDS (LIDS). The\nLIDS includes a sensor that is essentially an SNMP (Simple Network Man-\nagement Protocol) agent to retrieve data from the node’s MIB (management\ninformation base). The LIDS includes a file of signatures and performs mis-\nuse detection to detect attacks.\nInformation is shared among nodes by dispatching mobile agents, al-\nthough implementation details about this procedure are lacking. Also, the\nperformance and costs of the mobile agents have not been evaluated. Mo-\nbile agents have been studied for many years and proposed for fields such\nas network management and electronic commerce. However, the theoreti-\ncal advantages of mobile agents have been elusive.\nMobile agents have never seen much commercial success. Part of the\nreason is the need for universal adoption of a mobile agent platform (e.g.,\nJava Virtual Machine) which supports the execution and migration of mo-\nbile agents. Another reason is that mobile agents do not seem to perform\nany applications that static agents cannot. Finally, mobile agents introduce\nadditional security concerns because they involve the installation of new\n(possibly untrusted) code on a host. Special security mechanisms must be\ninstalled on hosts to ensure that mobile agents do not cause damage. Be-\ncause they require higher security, mobile agents are probably poor choices\nas a solution to security problems such as intrusion detection.\n4.4.9\nAODVSTAT\nAODVSTAT is an extension of STAT (state transition analysis technique) to\nintrusion detection in wireless networks that use AODV routing [32]. STAT\nis a stateful signature-based detection technique proposed earlier for wired\n" }, { "page_number": 172, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n161\nnetworks [33]. The premise is that computer attacks can be characterized\nas sequences of actions taken by an intruder. States represent a snapshot\nof a host’s volative, semi-permanent, and permanent memory.\nA complete representation of a successful attack starts from a safe ini-\ntial state, proceeds through a number of intermediate states, and ends in a\ncompromised state. States are characterized by assertions, which are func-\ntions with arguments returning Boolean values. These assertions describe\naspects of the security state of the system. Transitions between states are\nassociated with signature actions, which are actions by the intruder neces-\nsary for a successful attack. Omission of a signature action would prevent\nsuccessful completion of the attack.\nAODVSTAT applies the ideas of STAT to AODV-routed wireless net-\nworks. As mentioned earlier, AODV discovers routes on demand when a\npacket is ready to be sent. The source node floods a request through the\nnetwork, and a reply is returned by the destination or an intermediate node\nthat has a route to the destination. A malicious node could interfere with\nthe control packets of the routing protocol, or interfere with the forwarding\nof data packets.\nAODVSTAT sensors are placed on a subset of nodes for promiscuous\nsensing of radio channels. A sensor has two modes of operation. In stand-\nalone mode, a sensor looks for signs of attack only within its local neighbor-\nhood. In distributed mode, sensors periodically exchange “update” packets\ncontaining information about the neighboring nodes of each sensor. The\npurpose for sharing information is to detect attacks in a distributed way.\nAs in STAT, AODVSTAT works by stateful signature-based analysis of\nthe observed traffic. Each sensor has a file of attack signatures and looks\nfor a signature match with the traffic. A match triggers a response, usually\nan alert.\nAODVSTAT would have largely the same strengths and weaknesses as\nSTAT. As a misuse detection technique, AODVSTAT could accurately detect\ntypes of attacks that consist of sequential actions. A practical issue of how\nto update the attack signature files at all sensors in an ad hoc network\nhas not been addressed. Also, AODVSTAT has the same limitations as all\nmisuse detection techniques, i.e., the inability to detect attacks without an\nexisting signature. However, in a real implementation, it should be straight-\nforward to combine AODVSTAT with anomaly detection for the best of both\ntechniques.\n4.4.10\nTrust Model\nPirzada and McDonald [34] described an approach to building trust relation-\nships between nodes in an ad hoc network, but the method is essentially\nintrusion detection. It is assumed that nodes in the network passively mon-\nitor the packets received and forwarded by other nodes, called events.\n" }, { "page_number": 173, "text": "162\n■\nSecurity in Wireless Mesh Networks\nEvents are observed and given a weight, depending on the type of applica-\ntion requiring a trust relationship with other nodes. The weights reflect the\nsignificance of the observed event to the application. The trust values for all\nevents from a node are combined using weights to compute an aggregate\ntrust level for another node.\nTrust values could be viewed as link weights for the computation of\nroutes. Links with smaller weights would be links to more trusted nodes. A\nshortest-path routing algorithm would compute the most trustworthy paths.\nThe similarities between this scheme and previous IDS schemes are\nclear. Both approaches involve nodes observing the behavior of other\nnodes and making independent judgments about them. The only differ-\nence is that intrusion detection attempts to decide whether a node has been\ncompromised (misbehaving) or not, whereas Pirzada and McDonald’s trust\nmodel decides on the trustworthiness of a node.\n4.4.11\nRESANE\nRESANE (REputation-based Security in Ad hoc NEtworks) [35] takes a view\nsimilar to Pirzada and McDonald’s trust model. RESANE is not an IDS\nscheme per se, but uses intrusion detection techniques for a trust model.\nIt assumes that nodes are running an IDS scheme to identify nodes that\nare misbehaving. The problem addressed is how to make use of the IDS\ninformation.\nThe goal of RESANE is to calculate reputations for nodes and leverage\nreputations to motivate cooperation between nodes and good behavior\nthroughout the network. The idea is that a bad reputation will motivate a\nnode toward good behavior. If the node continues misbehavior, its repu-\ntation will continue to suffer and the node will become isolated from the\nrest of the network.\nA node calculates a reputation rating for a suspected neighbor from\nthe neighbor’s misbehaviors observed by the node. The node can also\ngather reputation ratings for that suspected neighbor from other neigh-\nboring nodes that have observed it. If a node detects a misbehavior by a\nsuspected neighbor, the node can proactively broadcast its information to\nother neighbors to help them protect themselves. Thus, the overall network\nis protected by cooperative information sharing.\n4.4.12\nCritical Nodes\nKarygiannis et al. advocated the concept of critical nodes [36]. These critical\nnodes are worth monitoring at the expense of more resources because\nthey have considerable effect on network performance. In other words, if\na critical node is malicious or misbehaving or fails, it would significantly\n" }, { "page_number": 174, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n163\ndegrade network performance. Non-critical nodes are not as important to\nmonitor when resources are limited (the usual case in ad hoc networks).\nThe notion of critical nodes may aid the problem of intrusion detection,\nbut the work does not address specifically how intrusions may be detected.\n4.4.13\nSCAN\nSCAN attempts to address two problems simultaneously: routing misbehav-\nior (control plane) and packet forwarding misbehavior (data plane) [37].\nRouting misbehavior is exhibited by a node that does not participate prop-\nerly in the routing protocol, e.g., false route advertisements. Packet for-\nwarding misbehavior refers to any intentional interference with the proper\nrelaying of packets, e.g., packet dropping and packet misrouting.\nSCAN is based on two central ideas that are similar to previous IDS\nschemes. First, each node monitors its neighbors independently. Different\nfrom a watchdog, which looks only for packet forwarding misbehavior,\nnodes in SCAN observe their neighbors for both routing misbehavior and\npacket forwarding misbehavior. The second idea is information cross vali-\ndation. Each node monitors its neighbors by cross-checking the overhead\ntransmissions with other nodes. Nodes in a neighborhood collaborate with\neach other through a distributed consensus protocol. A suspected node can\nbe eventually convicted of being malicious only after multiple neighbors\nhave reached that consensus. This assumes that the network density is suf-\nficiently high that a node can promiscuously overhear the packets sent and\nreceived by its neighbors, and nodes have multiple neighbors within range.\nFor routing misbehavior, SCAN requires two modifications to the usual\nAODV routing protocol. The usual routing update messages do not contain\nenough information for nodes to make judgments about routing misbehav-\nior. First, an additional field for “previous hop” is needed in route request\nmessages. Second, an additional field for “next hop” is needed in route re-\nply messages. This additional information in routing messages allows nodes\nto maintain part of the routing tables of its neighbors. The redundant rout-\ning information enables a node to examine the trustworthiness of future\nrouting updates from its neighbors.\nThe distributed consensus protocol is based on an “m out of N” algo-\nrithm, where N neighbors have been independently observing a suspected\nnode. The suspected node is convicted as malicious if at least m out of\nthe N nodes votes for that decision (based on observed misbehaviors).\nVarious strategies for choosing the value of m as a function of N are pro-\nposed: a fixed fraction of N, a constant value k, or a value depending on\na probability of correct detection and probability of false alarm.\nIf a node is convicted of being malicious, it is blocked from access to the\nnetwork. In SCAN, each node must present a valid token to interact with\n" }, { "page_number": 175, "text": "164\n■\nSecurity in Wireless Mesh Networks\nother nodes. Tokens for convicted nodes are revoked, and revoked tokens\nare tracked by each node by means of a token revocation list. Asymmetric\ncryptography is used to prevent forged tokens. Each token is signed by the\nsame secret key so it can be verified by a systemwide public key known\nto all nodes. Tokens are issued and renewed by a distributed algorithm. A\ntoken can be signed by a group of collaborating nodes, but not by a single\nnode. A token possessed by a node can be renewed by its neighbors if it\nexpires.\nSCAN has limitations and involves some overhead in terms of com-\nmunications and memory. The current SCAN scheme is limited to AODV,\nbut may be extended to other routing protocols if they are appropriately\nmodified (just as AODV messages must be modified with additional fields).\nAnother limitation of SCAN is a requirement for a dense ad hoc network\nbecause multiple neighbors must collaborate to form a consensual judg-\nment about a suspected node. Lastly, there is a requirement that collusion\namong attackers is limited.\n4.4.14\nDempster–Shafer\nChen and Venkataramanan [38] addressed the specific problem of combin-\ning the observations of multiple neighbors to form a consensual judgment\nabout a suspected node. Dempster–Shafer evidence theory [39] is proposed\nto be better than simple majority voting or a Bayesian approach. Essentially,\nDempster-Shafer theory allows observers to specify a level of uncertainty\nin their observation. In the context of intrusion detection, if each node has\na reputation or trustworthiness rating, that will be reflected by weighting\ntheir vote with a corresponding level of uncertainty. In other words, the\nvotes from untrusted nodes will be discounted, in comparison with votes\nfrom trusted nodes, in forming a consensual judgment.\n4.4.15\nOptimization of Limited Resources\nIn wireless networks, nodes may have limited resources to spend on in-\ntrusion monitoring and detection. On the other hand, intrusion detection\nis more effective when more traffic is monitored. The selection of nodes\nto operate IDS should consider the trade-off between detection efficiency\nand usage of limited resources. This trade-off was formulated as an integer\nlinear problem, where detection efficiency is maximized subject to a set of\nresource constraints [40].\nThe authors also considered a related problem where sensors could\nbe unreliable due to faults, power savings, or compromise [41]. Again,\nthe problem was formulated as an integer linear problem to minimize re-\nsource consumption subject to keeping a desired detection probability and\nthe possibility that sensors could be inactive.\n" }, { "page_number": 176, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n165\n4.5\nOpen Research Issues\nFor reasons mentioned earlier, intrusion detection is more difficult in wire-\nless mesh networks than wired networks. Intrusion detection continues to\nbe a difficult and open problem even in wired networks. In wired networks,\nit is relatively easy to collect traffic data, but the main challenge is detection\naccuracy. Neither of the two current analysis approaches, misuse detection\nor anomaly detection, is perfect. Fundamentally, misuse detection needs\nan attack signature to recognize an attack. New attacks without an existing\nsignature will be missed, resulting in a high rate of false negatives. Also, it\ntakes significant time to develop and distribute a new signature for a new\nattack. A new attack has a window of opportunity after its first detection\nwhere IDSs have not received a new signature yet. A new attack will not\nbe recognizable in the window of opportunity. Anomaly detection has a\ndifferent challenge: how to construct a normal behavior profile that will\nyield a low rate of false positives. Detection accuracy will continue to be\nthe main research issue in wireless mesh networks.\n4.5.1\nLack of Experience with Wireless Mesh Networks\nAnother open issue is the lack of experience with incidents in wireless\nmesh networks. In contrast, security incidents have been occurring in the\nInternet over the past 30 years. Although no comprehensive database of\nattacks exists, 30 years of experience have yielded a wealth of information\nabout Internet-based attacks. This wealth of information has helped the\nInternet security industry grow to considerable size, and a broad range of\nsecurity products are available.\nOn the other hand, wireless mesh networks are a recent development,\nand there is little real experience with security incidents. Attacks are mostly\nconjectured and theoretical at this point in time. Hence, it is really unknown\nhow to measure the progress or success of research. More real experience\nis needed, but will not be obtainable until wireless mesh networks are\ndeployed more widely in the field.\n4.5.2\nEvaluation Difficulties\nDifferent IDSs will detect and miss different attacks. A long-standing prob-\nlem has been how to fairly evaluate and compare different IDS. In the past,\nexperiments for wired networks have used test sets of various attacks and\nmeasured the detection rate. However, the results will obviously depend on\nthe types of attacks in the test set because different IDS methods will have\ndifferent strengths and weaknesses. Experimental comparisons of IDSs may\nalways be controversial. Also, considering the lack of experience with real\n" }, { "page_number": 177, "text": "166\n■\nSecurity in Wireless Mesh Networks\nwireless mesh networks, it is difficult to know what types of attacks will\nbe important or realistic.\n4.5.3\nIntrusion Tolerance\nAn indirectly related issue is the concept of intrusion tolerance. Intrusion\ndetection attempts to discover the occurrence of attacks and mostly leaves\nthe response to system administrators. Intrusion tolerance recognizes that\nattacks are inevitable and some attacks will be successful. The idea is to\ndesign networks from the beginning to maintain robust operation even in\nthe face of adversarial actions. For example, redundant paths can guarantee\nthat packets will still be delivered if an attacker brings down nodes. Clearly,\nintrusion tolerance is related to fault tolerance, except that fault tolerance\nassumes that faults are random and caused by equipment failures. Intru-\nsion tolerance assumes an intelligent attacker capable of strategic actions.\nIntrusion tolerance for wireless mesh networks is virtually unexplored.\n4.6\nConclusion\nThis chapter has reviewed the basic concepts of intrusion detection and\nsurveyed a number of proposals for intrusion detection in wireless mesh\nnetworks. The proposals are mostly for MANETs because wireless mesh\nnetworks are a relatively recent development, but the intrusion detection\nschemes are directly relevant to wireless mesh networks.\nA common theme in the research is the notion that nodes should in-\ndependently and concurrently monitor their local neighborhoods. This is\na necessity due to the decentralized nature of wireless mesh networks. A\nsecond common theme is the combination of observations from multiple\nnodes to form a consensual judgment about a suspected node. With these\ncommon themes, the various proposed intrusion detection schemes differ\nmainly in their details and not in their ideas.\nAt this point, a number of things are clear about the future of intrusion\ndetection. First, there is much room for improvement. The primary mea-\nsure of effective intrusion detection is low false positives and false nega-\ntives. This “proof” has not been convincingly offered by any scheme so far.\nSecond, the challenges imposed by wireless mesh networks imply that the\nintrusion detection problem will continue to be open for the foreseeable\nfuture. Finally, breakthrough progress may not be expected until wireless\nmesh networks are deployed more widely in the field. At this time, attacks\nand therefore intrusion detection are largely speculative and theoretical.\nMore real experience with wireless mesh networks will certainly help to\ncatalyze research progress.\n" }, { "page_number": 178, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n167\nReferences\n[1]\nS. McClure, J. Scambray, and G. Kurtz, Hacking exposed, 3rd ed., McGraw-\nHill, 2001.\n[2]\nL. Zhou and Z. Haas, Securing ad hoc networks, IEEE Network, vol. 13,\nNovember/December 1999, pp. 24–30.\n[3]\nH. Deng, W. Li, and D. Agrawal, Routing security in wireless ad hoc net-\nworks, IEEE Communications Magazine, vol. 40, October 2002, pp. 70–75.\n[4]\nK. Sanzgiri et al., Authenticated routing for ad hoc networks, IEEE J. on\nSel. Areas in Commun., vol. 23, March 2005, pp. 598–610.\n[5]\nN. Salem and J-P. Hubaux, Securing wireless mesh networks, IEEE Wireless\nCommunications, vol. 13, April 2006, pp. 50–55.\n[6]\nC. Basile, Z. Kalbarczyk, and R. Iyer, Neutralization of errors and attacks in\nwireless ad hoc networks, Int. Conf. on Dependable Systems and Networks\n(DSN), 2005, pp. 518–527.\n[7]\nN. Milanovic, M. Malek, A. Davidson, and V. Milutinovic, Routing and\nsecurity in mobile ad hoc networks, Computer, vol. 37, February 2004,\npp. 61–65.\n[8]\nH.\nYang\net\nal.,\nSecurity\nin\nmobile\nad\nhoc\nnetworks:\nChallenges\nand solutions, IEEE Wireless Communications, vol. 11, February 2004,\npp. 38–47.\n[9]\nY-C. Hu, A. Perrig, and D. Johnson, Wormhole attacks in wireless net-\nworks, IEEE J. on Sel. Areas in Communications, vol. 24, February 2006,\npp. 370–380.\n[10]\nJ. McHugh, Intrusion and intrusion detection, Int. J. of Information Security,\nvol. 1, August 2001, pp. 14–35.\n[11]\nS. Axelsson, Intrusion Detection Systems: A Survey and Taxonomy,\nTechnical report 99–15, Department of Computer Engineering, Chalmers\nUniversity of Technology, Sweden, March 2000.\n[12]\nR. Bace, Intrusion Detection, MacMillan Technical Publishing, 2000.\n[13]\nD. Marchette, Computer Intrusion Detection and Network Monitoring: A\nStatistical Viewpoint, Springer-Verlag, 2001.\n[14]\nR. Bejtlich, The Tao of Network Security Monitoring: Beyond Intrusion\nDetection, Addison-Wesley, 2005.\n[15]\nS. Northcutt and J. Novak, Network Intrusion Detection, 3rd ed., Pearson\nEducation, 2003.\n[16]\nS. Northcutt, M. Cooper, M. Fearnow, and K. Frederick, Intrusion Signatures\nand Analysis, New Riders Publishing, 2001.\n[17]\nK. Cox and C. Gerg, Snort and IDS Tools, O’Reilly Media, 2004.\n[18]\nR. Bruno, M. Conti, and E. Gregori, Mesh networks: Commodity multihop\nad hoc networks, IEEE Communications Magazine, vol. 43, March 2005,\npp. 123–131.\n[19]\nM. Lee, J. Zheng, Y-G. Ko, and D. Shrestha, Emerging standards for wire-\nless mesh networks, IEEE Wireless Communications, vol. 13, April 2006,\npp. 56–63.\n[20]\nI. Akyildiz, X. Wang, and W. Wang, Wireless mesh networks: A survey,\nComputer Networks, vol. 47, 2005, pp. 445–487.\n" }, { "page_number": 179, "text": "168\n■\nSecurity in Wireless Mesh Networks\n[21]\nI. Akyildiz, X. Wang, A survey on wireless mesh networks, IEEE Commu-\nnications Magazine, vol. 43, September 2005, pp. S23–S30.\n[22]\nA. Mishra, K. Nadkarni, and A. Patcha, Intrusion detection in wireless ad\nhoc networks, IEEE Wireless Communications, vol. 11, February 2004, pp.\n48–60.\n[23]\nK. Bradley et al., Detecting disruptive routers: A distributed network\nmonitoring approach, IEEE Network, vol. 12, September/October 1998,\npp. 50–60.\n[24]\nY. Zhang and W. Lee, Intrusion Detection in Wireless Ad-hoc Networks,\n6th Annual ACM Int. Conf. on Mobile Computing and Networking, Boston,\n2000, pp. 275–283.\n[25]\nY. Zhang, W. Lee, and Y-A. Huang, Intrusion detection techniques for\nmobile wireless networks, Wireless Networks, vol. 9, 2003, pp. 545–556.\n[26]\nS. Marti, T. Giuli, K. Lai, and M. Baker, Mitigating Routing Misbehavior in\nMobile Ad hoc Networks, 6th Annual ACM Int. Conf. on Mobile Computing\nand Networking, Boston, 2000, pp. 255–265.\n[27]\nR. Ramanujan, A. Ahamad, J. Bonney, R. Hagelstrom, and K. Thurber,\nTechniques for Intrusion-resistant Ad hoc Routing Algorithms (TIARA),\nIEEE MILCOM 2000, Los Angeles, 2000, pp. 660–664.\n[28]\nS. Bhargava and D. Agrawal, Security Enhancements in AODV Protocol\nfor Wireless Ad hoc Networks, 2001 IEEE Vehicular Technology Conf.\n(VTC 2001), 2001, pp. 2143–2147.\n[29]\nS. Buchegger and J-Y. Le Boudec, Performance Analysis of the CONFIDANT\nProtocol (Cooperation of Nodes: Fairness in Dynamic Ad-hoc Networks),\n3rd ACM Int. Symp. on Mobile Ad hoc Networks and Computing,\nSwitzerland, 2002, pp. 226–236.\n[30]\nF. Kargl, A. Klenk, M. Weber, and S. Schlott, Sensors for Detection of\nMisbehaving Nodes in MANETs, Detection of Intrusion and Malware and\nVulnerability Assessment (DIMVA 2004), Dortmund, Germany, 2004.\n[31]\nR. Puttini, J-M. Percher, L. Me, and R. de Sousa, A Fully Distributed IDS for\nMANET, 9th Int. Symp. on Computers and Commun. (ISCC 2004), 2004,\npp. 331–338.\n[32]\nG. Vigna et al., An Intrustion Detection Tool for AODV-based Ad hoc\nWireless Networks, Annual Computer Security Applications Conf. (ACSAC\n2004), Tucson, 2004, pp. 16–27.\n[33]\nK. Ilgun, R. Kemmerer, and P. Porras, State transition analysis: A rule-based\nintrusion detection approach, IEEE Trans. on Software Engineering, vol. 21,\n1995, pp. 181–199.\n[34]\nA. Pirzada and C. McDonald, Establishing Trust in Pure Ad hoc Networks,\n27th Australian Conf. on Computer Science, Dunedin, New Zealand, 2004,\npp. 47–54.\n[35]\nY. Rebahi, V. Mujica, and D. Sisalem, A Reputation-based Trust Mechanism\nfor Ad hoc Networks, 10th IEEE Symp. on Computers and Communications\n(ISCC 2005), 2005, pp. 37–42.\n" }, { "page_number": 180, "text": "Intrusion Detection in Wireless Mesh Networks\n■\n169\n[36]\nA. Karygiannis, E. Antonakakis, and A. Apostolopoulos, Detecting Critical\nNodes for MANET Intrusion Detection, 2nd Int. Workshop on Security,\nPrivacy, and Trust in Pervasive and Ubiquitous Computing (SecPerU 2006),\n2006, pp. 7–15.\n[37]\nH. Yang, J. Shu, X. Meng, and S. Lu, SCAN: Self-organized network-layer se-\ncurity in mobile ad hoc networks, IEEE J. on Sel. Areas in Communications,\nvol. 24, February 2006, pp. 261–273.\n[38]\nT. Chen and V. Venkataramanan, Dempster–Shafer theory for intrusion de-\ntection in ad hoc networks, IEEE Internet Computing, vol. 9, November/\nDecember 2005, pp. 35–41.\n[39]\nG. Shafer, A Mathematical Theory of Evidence, Princeton University Press,\n1976.\n[40]\nD. Subhadrabandhu, S. Sarkar, and F. Anjum, A framework for misuse de-\ntection in ad hoc networks — Part I, IEEE J. on Sel. Areas in Communi-\ncations, vol. 24, February 2006, pp. 274–289.\n[41]\nD. Subhadrabandhu, S. Sarkar, and F. Anjum, A framework for misuse de-\ntection in ad hoc networks — Part II, IEEE J. on Sel. Areas in Communi-\ncations, vol. 24, Feb. 2006, pp. 290–304.\n" }, { "page_number": 181, "text": "" }, { "page_number": 182, "text": "Chapter 5\nSecure Routing in\nWireless Mesh Networks\nManel Guerrero Zapata\nContents\n5.1\nIntroduction ........................................................ 172\n5.2\nRelated Work ....................................................... 172\n5.3\nDesigning a Secure Routing Protocol.............................. 175\n5.4\nSecurity Requirements.............................................. 176\n5.5\nSecuring Wireless Mesh Network Routing Protocols.............. 177\n5.6\nSecuring Ad hoc Network Routing Protocols ..................... 178\n5.7\nAd hoc On-Demand Vector Routing............................... 179\n5.8\nSecurity Flaws of AODV ........................................... 181\n5.9\nSecure Ad hoc On-Demand Distance Vector ...................... 182\n5.9.1\nSAODV Hash Chains....................................... 183\n5.9.2\nSAODV Digital Signatures ................................. 184\n5.9.3\nSecuring Error Messages ................................... 186\n5.9.4\nPersistence of Sequence Numbers ........................ 187\n5.10\nOpen Issues ........................................................ 187\n5.11\nAODV Message Formats ........................................... 188\n5.12\nSecure AODV Extensions .......................................... 190\nReferences................................................................. 193\nMost routing protocols for client wireless mesh networks (WMNs) were\ndesigned without having security in mind. In most of their specifications it\nis assumed that all the nodes in the network are friendly. The security issue\nwas postponed and there used to be the common feeling that it would be\n171\n" }, { "page_number": 183, "text": "172\n■\nSecurity in Wireless Mesh Networks\npossible to make those routing protocols secure by retrofitting pre-existing\ncryptosystems.\nNevertheless, securing network transmissions without securing the rout-\ning protocols is not sufficient. Unless fixed networks (where one might\nassume that routers are trusted nodes) in a wireless network (where all\nthe nodes are also routing nodes) are secure, malicious nodes might attack\nrouting protocols to impersonate other nodes and inject forged routing in-\nformation. Moreover, by retrofitting cryptosystems (like IPSec [19]) security\nis not necessarily achieved.\nTherefore, in client WMNs with security needs, there must be two secu-\nrity systems: one to protect the data transmission and one to make the rout-\ning protocol secure. There are already well-studied, point-to-point security\nsystems that can be used for protecting network transmissions. But there\nwas not much work to make wireless routing protocols discover routes in\na secure manner [18,31,37] until recently.\n5.1\nIntroduction\nSome aspects of wireless and ad hoc networks have interesting security\nproblems [2,33,37]. Routing is one such aspect. Several routing protocols\nfor these kind of networks have been developed, particularly in the MANET\nWorking Group of the Internet Engineering Task Force (IETF). Surveys of\nrouting protocols for ad hoc wireless networks are presented in [29,30] and,\nmore recently, in [15] and [34].\n5.2\nRelated Work\nBy the year 2000 there was very little published work on the security issues\nin ad hoc and wireless network routing protocols. Neither the survey by\nRamanathan and Steenstrup [29] in 1996, nor the survey by Royer and\nToh [30] in 1999 mention security. None of the draft proposals in the\nIETF MANET Working Group had a non-trivial “security considerations”\nsection. Actually, most of them assumed that all the nodes in the network\nare friendly, and a few declare the problem out-of-scope by assuming some\ncanned solution like IPSec may be applicable.\nSecurity issues with routing in general have been addressed by sev-\neral researchers (e.g., [13,32]) at the end of the 20th century. And, later,\nsome work has been done to secure ad hoc networks by using misbehav-\nior detection schemes (e.g., [23]). This approach has two main problems:\nfirst, it is quite likely that it will not be feasible to detect several kinds of\nmisbehavior (especially because it is very hard to distinguish misbehaving\n" }, { "page_number": 184, "text": "Secure Routing in Wireless Mesh Networks\n■\n173\nfrom transmission failures and other kind of failures); and second, it has\nno real means to guarantee the integrity and authentication of the routing\nmessages.\nHash chains had being used as an efficient way to obtain authentication\nin several approaches that tried to secure routing protocols. In [5,13,28]\nthey use them to provide delayed key disclosure. In [36], hash chains\nare used to create one-time signatures that can be verified immediately.\nThe main drawback of all the above approaches is that they require clock\nsynchronization.\nIn their paper on securing ad hoc networks [37] in 1999, Zhou and Haas\nprimarily discuss key management. They devote a section to secure routing,\nbut essentially conclude that “nodes can protect routing information in\nthe same way they protect data traffic.” They also observe that denial-\nof-service attacks against routing will be treated as damage and routed\naround.\nDahill et al. [7] proposed ARAN in 2001, a routing protocol for ad hoc\nnetworks that uses authentication and requires the use of a trusted cer-\ntificate server. In ARAN, every node that forwards a route discovery or a\nroute reply message must also sign it (which is very computing-power-\nconsuming and causes the size of the routing messages to increase at each\nhop). In addition, it is prone to reply attacks using error messages unless\nthe nodes have time synchronization.\nIn October 2001, the first draft of SAODV [10] was sent to the MANET\nmailing list. SAODV [11,12] is an extension of the AODV routing proto-\ncol that can be used to protect the route discovery mechanism providing\nsecurity features like integrity and authentication, and it only requires orig-\ninators of routing messages to sign the routing messages (as opposed to\nARAN, in which all the forwarding nodes sign the messages).\nIn 2002, Papadimitratos and Haas [27] proposed a protocol (SRP) that\ncan be applied to several existing routing protocols (in particular DSR [17]).\nSRP requires that, for every route discovery, source and destination must\nhave a security association between them. Furthermore, the paper does\nnot even mention route error messages. Therefore, they are not protected,\nand any malicious node can just forge error messages with other nodes as\nsource.\nIn SEAD [16], hash chains are also used in combination with DSDV-\nSQ [3] (this time to authenticate hop counts and sequence numbers). At\nevery given time each node has its own hash chain. The hash chain is\ndivided into segments; elements in a segment are used to secure hop counts\nin a way similar to SAODV. The size of the hash chain is determined when\nit is generated. After using all the elements of the hash chain, a new one\nmust be computed.\nSEAD can be used with any suitable authentication and key distribution\nscheme. But finding such a scheme is not straightforward.\n" }, { "page_number": 185, "text": "174\n■\nSecurity in Wireless Mesh Networks\nAriadne [16] is based on DSR [17] and TESLA [27] (on which its authen-\ntication mechanism is based). It also requires clock synchronization, which\nis, arguably, an unrealistic requirement for ad hoc networks.\nIn principle, the same approach that SAODV takes to protect AODV\ncould be used to create a “secure version” of other routing protocols: sign-\ning the non-mutable routing information by the node to which the route\nwill be processed, and securing the hop count by hash chains. In case\nthere are some other mutable fields, how to protect each of them should\nbe studied.\nNevertheless, if the routing protocol has some other mutable informa-\ntion than the hop count (and it does not mutate in a predictable way),\nprotecting this information might end up being quite complex. It will prob-\nably require that the intermediate nodes that mutate part of the message\nalso have to sign it. This will, typically, imply a reduction of performance\n(due to all the additional cryptographic computations) and also a possible\ndecrease of the overall security.\nIf the routing protocol to be secured is DSR for mobile ad hoc networks\n[17], then the main problem will be that DSR includes in its routing messages\nthe IP addresses of all the intermediate nodes that have forwarded the\npacket.\nIntermediate nodes could sign the routing message after adding its own\nIP address, and verify all the signatures in every routing message. But this\nwould greatly decrease the performance of the routing discovery, and it is\nnot really worthwhile, taking into account that the routes to the intermediate\nnodes are going to be used very seldom. Anyway, hash chains should be\nused to avoid that a malicious node would eliminate intermediate nodes\nand their signatures from the routing message (a very similar technique is\nalso used in [16]).\nAnother solution would be that intermediate nodes would sign the rout-\ning message, but that a node would only verify the signature of an interme-\ndiate node when it needs to send a packet to this route. But it still requires\nall intermediate nodes to sign the message (which is not good when the\nmessage is a route request).\nTherefore, maybe a better solution would be that intermediate nodes do\nnot sign the message. Later on, if a node that received that routing message\nwants to use a route to one of those intermediate nodes, it should request\na signature from the intermediate node with a unicast message.\nObviously, a much more detailed analysis should be made to study\nthe different attacks that can be performed against DSR and against this\n“secure DSR” to see if there are new attacks as a consequence of differences\nbetween AODV and DSR.\nSRP [24] and Ariadne [16] also attempt to secure DSR. Nevertheless, SRP\nrequires that, for every route discovery, source and destination must have\n" }, { "page_number": 186, "text": "Secure Routing in Wireless Mesh Networks\n■\n175\na security association between them, and does not protect error messages.\nAriadne requires clock synchronization, which can be considered unreal-\nistic for ad hoc networks.\nMore recently and more focused on mesh networks, a paper by\nAsherson and Hutchison [1] has as a starting point the concern that routing\nalgorithms designed for ad hoc networks might not be applied straight-\nforward to WMNs. Nevertheless, it concludes giving as a solution to use\ndifferent routing protocols for the infrastructure part and for the ad hoc\npart (which would use a routing protocol for ad hoc networks), therefore\nadopting the same approach as the one used in the Internet.\nIn the area of routing metrics for mesh networks, Yang, Wang, and\nKravets [35] have studied how the use of different routing metrics affects\nthe performance of the routing protocol in mesh networks. Nevertheless,\nthey leave as future work the problem of how to transmit routing metrics\nin a secure manner.\nFinally, the recent standardization efforts of the IEEE 802.11s (the IEEE\nstandard for mesh networking) are considering MANET routing protocols\nlike AODV [25,26] and OLSR [6] as their mesh routing protocols, as noted\nin the performance comparison paper by Chen, Lee, Maniezzo, and\nGerla [4].\n5.3\nDesigning a Secure Routing Protocol\nWhen designing a secure routing protocol, as with any secure protocol,\nthings need to be kept as simple and neat as possible, so they can be\nproperly analyzed.\nFerguson and Schneier, in their paper “A Cryptographic Evaluation of\nIPsec” [8], conclude that the complexity of IPsec results in inefficiencies\nand weaknesses which make it weaker and very hard to analyze how\nsecure it is. The bottom line is that creating a too-complex solution makes\nit unfeasible to verify if it is a good solution.\nTo keep the design of a secure routing protocol as neat as possible, it\nis convenient to make a clear distinction of the following items:\n■\nThe scenario (or scenarios) it is going to protect\n■\nThe security features that this scenario requires\n■\nThe security mechanisms that will fulfill those security features\nOnce the design of the secure routing protocol is done, it is time to\nanalyze whether it indeed works, and, because the three items listed above\nare clearly separated in the design, it is much easier to perform such analysis\nbecause it can be split into the following parts:\n" }, { "page_number": 187, "text": "176\n■\nSecurity in Wireless Mesh Networks\n■\nThe analysis of requirements: Whether the security features are\nenough for the targeted scenario.\n■\nThe analysis of mechanisms: Whether the security mechanisms are\nindeed fulfilling all the security requirements. When doing this, it\nwill be found that there are still some attacks that can be performed\nagainst your system. Some of them typically will not be completely\navoided because of a trade-off between security and feasibility.\n■\nThe analysis of feasibility: Whether the security mechanisms have\nrequirements that are not feasible in the targeted scenario.\n5.4\nSecurity Requirements\nIn most domains, the primary security service is authorization. Routing is\nno exception. Typically, a router needs to make two types of authorization\ndecisions. First, when a routing update is received from the outside, the\nrouter needs to decide whether to modify its local routing information base\naccordingly. This is import authorization. Second, a router may carry out\nexport authorization whenever it receives a request for routing information.\nImport authorization is the critical service.\nIn traditional routing systems, authorization is a matter of policy. For ex-\nample, gated, a commonly used routing program,1 allows the administrator\nof a router to set policies about whether and how much to trust routing\nupdates from other routers, e.g., statements like “trust router X about routes\nto networks A and B.” In mobile wireless networks, such static policies are\nnot sufficient (and unlikely to be relevant).\nAuthorization requires other security services such as authentication\nand integrity. Techniques like digital signatures and message authentica-\ntion codes are used to provide these services.\nIn the context of routing, confidentiality and non-repudiation are not\nnecessarily critical services [13]. Zhou and Haas [37] argue that non-\nrepudiation is useful in an ad hoc network for isolating misbehaving routers:\na router A which received an “erroneous message” from another router B\nmay use this message to convince other routers that B is misbehaving. This\nwould indeed be useful if there is a reliable way of detecting erroneous\nmessages. This does not appear to be an easy task.\nThe problem of compromised nodes is not addressed here because\nit would probably require some sort of mechanism to allow the owner\nto confirm its presence. Availability is considered to be outside of scope.\nAlthough of course it would be desirable, it does not seem to be feasible to\nprevent denial-of-service attacks in a network that uses wireless technology\n1 http://www.gated.org\n" }, { "page_number": 188, "text": "Secure Routing in Wireless Mesh Networks\n■\n177\n(where an attacker can focus on the physical layer without bothering to\nstudy the routing protocol).\nTherefore, in this research work the following requirements were con-\nsidered:\n■\nImport authorization: It is important to note that this is not referring\nto the traditional meaning of authorization. What it means is that\nthe ultimate authority about routing messages regarding a certain\ndestination node is that node itself. Therefore, route information\nwill only be authorized in a routing table if that route information\nconcerns the node that is sending the information. In this way, if a\nmalicious node lies about it, the only thing it will cause is that others\nwill not be able to route packets to the malicious node.\n■\nSource authentication: Nodes need to be able to verify that the node\nis the one it claims to be.\n■\nIntegrity: In addition, nodes need to be able to verify that the routing\ninformation has arrived unaltered.\n■\nThe two last security services combined build data authentication,\nand they are requirements derived from the import authorization\nrequirement.\nFinally, it is quite likely that, for a small team of nodes that trust each\nother and that want to create an ad hoc network where the messages are\nonly routed by members of the team, the simplest way to keep secret their\ncommunications is to encrypt all messages (routing and data) with a “team\nkey.” Every member of the team would know the key and, therefore, it\nwould be able to encrypt and decrypt every single packet. Nevertheless, this\ndoes not scale well and the members of the team have to trust each other.\nSo it can be used only for a very small subset of the possible scenarios.\nThat renders asymmetric cryptography as the most suitable option for most\nwireless scenarios.\n5.5\nSecuring Wireless Mesh Network\nRouting Protocols\nIf we agree with the idea reflected in the paper by Asherson and Hutchison\n[1], that the best approach is to use different routing protocols for the\ninfrastructure part and for the ad hoc part (which would use a routing\nprotocol for ad hoc networks), then the problem of securing WMN routing\nprotocols becomes a much simpler one. The mesh network is composed\nby the infrastructure part and by the ad hoc networks that are connected\nto the infrastructure network through the access points.\n" }, { "page_number": 189, "text": "178\n■\nSecurity in Wireless Mesh Networks\nThe infrastructure part can use a routing protocol suitable for fixed\nnetworks, the ad hoc networks can use a secure routing protocol suitable\nfor MANET networks, and the access points play as gateways of both the\ninfrastructure and the ad hoc networks.\nBecause the access points act as gateways between two networks that\nuse different routing protocols, they will use “administrative distances” to\nprioritize the use of routes of the infrastructure part. Remember that, in case\nthere is a route to the same destination provided by two different routing\nprotocols, the one with lowest “administrative distance” is used.\nRouting protocols for fixed networks are relatively easy to secure. There-\nfore, the real challenge is to secure the routing protocol of the ad hoc part\nof the mesh network.\n5.6\nSecuring Ad hoc Network Routing Protocols\nIn an ad hoc network, from the point of view of a routing protocol, there\nare two kinds of messages: the routing messages and the data messages.\nThe routing protocol uses routing messages to establish the routes that are\nneeded to transmit data messages, and, in the case of a reactive routing\nprotocol, it sees the data messages and refreshes the lifetimes of the routes\nthat those data messages use.\nThe two kinds of messages are different in nature and security needs.\nData messages are end-to-end and can be protected with any end-to-end\nsecurity system (like IPSec). On the other hand, routing messages are sent\nto neighbors, processed, possibly modified, and re-sent. Moreover, as a\nresult of the processing of the routing message, a node might modify its\nrouting table. This creates the need for the intermediate nodes to be able\nto authenticate the information contained in the routing messages (a need\nthat does not exist in end-to-end communications) to be able to apply their\nimport authorization policy.\nAnother consequence of the nature of the transmission of routing mes-\nsages is that, in many cases, there will be some parts of those messages\nthat will change during their propagation. This is very common in distance-\nvector routing protocols, where the routing messages usually contain a hop\ncount of the route they are requesting or providing. Therefore, in a routing\nmessage, two types of information could be distinguished: mutable and\nnon-mutable. It is desired that the mutable information in a routing mes-\nsage is secured in such a way that no trust in intermediate nodes is needed.\nOtherwise, securing the mutable information will be much more expensive\nin computation, plus the overall security of the system will greatly decrease.\nIf the security system being used to secure the data messages in a wire-\nless network is IPSec, it is necessary that the IPSec implementation can use\nas a selector the TCP and UDP port numbers. This is because it is necessary\n" }, { "page_number": 190, "text": "Secure Routing in Wireless Mesh Networks\n■\n179\nthat the IPSec policy will be able to apply certain security mechanisms to\nthe data packets and just bypass the routing packets (that can be identified\nbecause they use a reserved Transport layer port number).\n5.7\nAd hoc On-Demand Vector Routing\nThe Ad hoc On-Demand Vector Routing (AODV) protocol [25,26] is a reac-\ntive routing protocol for ad hoc and mobile networks that maintains routes\nonly between nodes which need to communicate. The routing messages\ndo not contain information about the whole route path, but only about the\nsource and the destination. Therefore, routing messages do not have an\nincreasing size. It uses destination sequence numbers to specify how fresh\na route is (in relation to another), which is used to grant loop freedom.\nWhenever a node needs to send a packet to a destination for which\nit has no “fresh enough” route (i.e., a valid route entry for the destination\nwhose associated sequence number is at least as great as the ones contained\nin any RREQ that the node has received for that destination), it broadcasts\na route request (RREQ) message to its neighbors. Each node that receives\nthe broadcast sets up a reverse route toward the originator of the RREQ,\nunless it has a “fresher” one (Figure 5.1).\nWhen the intended destination (or an intermediate node that has a\n“fresh enough” route to the destination) receives the RREQ, it replies by\nRoute request broadcast (S D)\nReverse routes after the broadcast\nS\nA\nB\nC\nD\nF\nE\nG\nH\nF\nE\nS\nA\nB\nC\nD\nH\nG\nFigure 5.1\nRoute Request. After the RREQ broadcast, D has in its routing table that\nthe next hop to S is D. The rest of the nodes also have in their routing table which is\nthe next hop to S.\n" }, { "page_number": 191, "text": "180\n■\nSecurity in Wireless Mesh Networks\nRoutes after the route reply\nS\nA\nB\nC\nD\nF\nE\nG\nH\nF\nE\nS\nA\nB\nC\nD\nH\nG\nRoute replies\nFigure 5.2\nRoute Reply. After S receives the RREP, all the nodes between S and D\nknow which are the next hops to S and D. The rest of the nodes (E, F, G, and H) also\nhave in their routing table which is the next hop to S. If they do not use that route,\nit will expire.\nsending a Route Reply (RREP). It is important to note that the only mutable\ninformation in an RREQ and in an RREP is the hop count (which is being\nmonotonically increased at each hop). The RREP is unicast back to the\noriginator of the RREQ (Figure 5.2). At each intermediate node, a route to\nthe destination is set (again, unless the node has a “fresher” route than the\none specified in the RREP). In the case that the RREQ is replied to by an\nintermediate node (and if the RREQ had set this option), the intermediate\nnode also sends an RREP to the destination. In this way, it can be granted\nthat the route path is being set up bidirectionally. In the case that a node\nreceives a new route (by an RREQ or by an RREP) and the node already\nhas a route “as fresh” as the received one, the shortest one will be updated.\nIf there is a subnet (a collection of nodes identified by a common net-\nwork prefix) that does not use AODV as its routing protocol and wants to\nbe able to exchange information with an AODV network, one of the nodes\nof the subnet can be selected as the “network leader.” The network leader\nis the only node of the subnet that sends, forwards, and processes AODV\nrouting messages. In every RREP that the leader issues, it sets the prefix\nsize of the subnet.\nOptionally, a Route Reply Acknowledgment (RREP-ACK) message may\nbe sent by the originator of the RREQ to acknowledge the receipt of the\nRREP. An RREP-ACK message has no mutable information.\nIn addition to these routing messages, a Route Error (RERR) message is\nused to notify the other nodes that certain nodes are not reachable anymore\n" }, { "page_number": 192, "text": "Secure Routing in Wireless Mesh Networks\n■\n181\nRoutes after the failure\nS\nA\nB\nC\nD\nF\nE\nG\nH\nF\nE\nS\nA\nB\nC\nD\nH\nG\nRoutes before the failure\nLink failure\nFigure 5.3\nRoute Error. When a link failure is detected, all the nodes between S and\nD get notified about it by a Route Error (RERR) message and erase their routes to\nS and D.\ndue to a link breakage (Figure 5.3). When a node rebroadcasts an RERR,\nit only adds the unreachable destinations to which the node might forward\nmessages. Therefore, the mutable information in an RERR is the list of\nunreachable destinations and the counter of unreachable destinations in-\ncluded in the message. It is predictable that, at each hop, the unreachable\ndestination list may not change or become a subset of the original one.\n5.8\nSecurity Flaws of AODV\nBecause AODV has no security mechanisms, malicious nodes can perform\nmany attacks just by not behaving according to the AODV rules. A malicious\nnode M can carry out the following attacks (among many others) against\nAODV:\n1.\nImpersonate a node S by forging an RREQ with its address as the\noriginator address.\n2.\nWhen forwarding an RREQ generated by S to discover a route to D,\nreduce the hop count field to increase the chances of being in the\nroute path between S and D so it can analyze the communica-\ntion between them. A variant of this is to increment the destination\nsequence number to make the other nodes believe that this is a\n“fresher” route.\n" }, { "page_number": 193, "text": "182\n■\nSecurity in Wireless Mesh Networks\n3.\nImpersonate a node D by forging an RREP with its address as a\ndestination address.\n4.\nImpersonate a node by forging an RREP that claims that the node is\nthe destination and, to increase the impact of the attack, claims to\nbe a network leader of the subnet S N with a big sequence number\nand send it to its neighbors. In this way it will become (at least\nlocally) a black hole for the whole subnet S N.\n5.\nSelectively, not forward certain RREQs and RREPs, not reply to cer-\ntain RREPs, and not forward certain data messages. This kind of\nattack is especially hard even to detect because transmission errors\nhave the same effect.\n6.\nForge an RERR message pretending it is the node S and send it\nto its neighbor D. The RERR message has a very high destination\nsequence number dsn for one of the unreachable destinations (U ).\nThis might cause D to update the destination sequence number\ncorresponding to U with the value dsn and, therefore, future route\ndiscoveries performed by D to obtain a route to U will fail (because\nU ’s destination sequence number will be much smaller than the one\nstored in D’s routing table).\n7.\nAccording to the AODV specification [25], the originator of an RREQ\ncan put a much bigger destination sequence number than the real\none. In addition, sequence numbers wrap around when they reach\nthe maximum value allowed by the field size. This allows a very\neasy attack, where an attacker is able to set the sequence number\nof a node to any desired value by just sending two RREQ messages\nto the node.\n5.9\nSecure Ad hoc On-Demand Distance Vector\nAssume that there is a key management sub-system that makes it possible\nfor each ad hoc node to obtain public keys from the other nodes of the\nnetwork. Further, each ad hoc node is capable of securely verifying the\nassociation between the identity of a given ad hoc node and the public\nkey of that node. How this is achieved depends on the key management\nscheme. Do not worry about how key management is achieved at this\npoint.\nSAODV uses two mechanisms to secure the AODV messages: digital\nsignatures (why we need the key management sub-system) to authenticate\nthe non-mutable fields of the messages, and hash chains to secure the\nhop count information (the only mutable information in the messages).\nFor the non-mutable information, authentication is performed in an end-\nto-end manner, but the same kind of techniques cannot be applied to the\nmutable information.\n" }, { "page_number": 194, "text": "Secure Routing in Wireless Mesh Networks\n■\n183\nThe information relative to the hash chains and the signatures is transmit-\nted with the AODV message as an extension message that will be referred\nto as Signature Extension.\n5.9.1\nSAODV Hash Chains\nSAODV uses hash chains to authenticate the hop count of RREQ and RREP\nmessages in such a way that allows every node that receives the message\n(either an intermediate node or the final destination) to verify that the hop\ncount has not been decremented by an attacker. This prevents an attack\nof type 2. A hash chain is formed by applying a one-way hash function\nrepeatedly to a seed.\nEvery time a node originates an RREQ or RREP message, it performs the\nfollowing operations:\n■\nGenerates a random number (seed).\n■\nSets the Max Hop Count field to the TimeToLive value (from the IP\nheader).\nMax Hop Count = TimeToLive\n■\nSets the Hash field to the seed value.\nHash = seed\n■\nSets the Hash Function field to the identifier of the hash function\nthat it is going to use. The possible values are shown in Table 5.1.\nHash Function = h\n■\nCalculates Top Hash by hashing seed Max Hop Count times.\nTop Hash = hMax Hop Count(seed)\nwhere h is a hash function, and hi(x) is the result of applying the\nfunction h to x i times.\nIn addition, every time a node receives an RREQ or RREP message, it\nperforms the following operations to verify the hop count:\n■\nApplies the hash function h Maximum Hop Count minus Hop Count\ntimes to the value in the Hash field, and verifies that the resultant\nvalue is equal to the value contained in the Top Hash field.\nTop Hash = hMax Hop Count−Hop Count(Hash)\nwhere a = b reads: to verify that a and b are equal.\n" }, { "page_number": 195, "text": "184\n■\nSecurity in Wireless Mesh Networks\nTable 5.1\nPossible Values of the Hash\nFunction Field\nValue\nHash Function\n0\nReserved\n1\nMD5HMAC96 [21]\n2\nSHA1HMAC96 [22]\n3–127\nReserved\n128–255\nImplementation dependent\n■\nBefore rebroadcasting an RREQ or forwarding an RREP, a node\napplies the hash function to the Hash value in the Signature\nExtension to account for the new hop.\nHash = h (Hash)\nThe Hash Function field indicates which hash function has to be used\nto compute the hash. Trying to use a different hash function will just\ncreate a wrong hash without giving any advantage to a malicious node.\nHash Function, Max Hop Count, Top Hash, and Hash fields are transmit-\nted with the AODV message in the Signature Extension, and as it will be\nexplained later, all of them but the Hash field are signed to protect its\nintegrity.\nFigure 5.4 shows the mechanisms to do the hash chain initialization,\nhop count verification, and hop count incrementation.\n5.9.2\nSAODV Digital Signatures\nDigital signatures are used to protect the integrity of the non-mutable data\nin RREQ and RREP messages. That means that they sign everything but\nthe Hop Count of the AODV message and the Hash from the SAODV\nextension.\nThe main problem in applying digital signatures is that AODV allows in-\ntermediate nodes to reply to RREQ messages if they have a “fresh enough”\nroute to the destination. While this makes the protocol more efficient, it\nalso makes it more complicated to secure. The problem is that an RREP\nmessage generated by an intermediate node should be able to sign it on\nbehalf of the final destination; in addition, it is possible that the route stored\nin the intermediate node would be created as a reverse route after receiv-\ning an RREQ message (which means that it does not have the signature for\nthe RREP).\nTo solve this problem, SAODV offers two alternatives. The first one (and\nalso the obvious one) is that, if an intermediate node cannot reply to an\nRREQ message because it cannot properly sign its RREP message, it just\n" }, { "page_number": 196, "text": "Secure Routing in Wireless Mesh Networks\n■\n185\nHash = h(Hash)\nHop count incrementation\nHop_Count = Hop_Count + 1\nTop_Hash = hMax_Hop_Count(seed)\n hMax_Hop_Count(hash)\n= Top_Hash?\nHash = seed\nMax_Hop_Count = TimeToLive\nGenerate seed\nHash chain initialization\nHop count verification\nVerification\nfailed\nNo\nYes\nHop count\nverified\nFigure 5.4\nProtection of the hop count through hash chains.\nbehaves as if it did not have the route and forwards the RREQ message.\nThe second is that, every time a node generates an RREQ message, it also\nincludes the RREP flags, the prefix size, and the signature that can be used\n(by any intermediate node that creates a reverse route to the originator of\nthe RREQ) to reply to an RREQ that asks for the node that originated the first\nRREQ. Moreover, when an intermediate node generates an RREP message,\nthe lifetime of the route has changed from the original one. Therefore, the\nintermediate node should include both lifetimes (the old one is needed\nto verify the signature of the route destination) and sign the new lifetime.\nIn this way, the original information of the route is signed by the final\ndestination and the lifetime is signed by the intermediate node.\nTo distinguish the different SAODV extension messages, the ones that\nhave two signatures are called RREQ and RREP Double Signature\nExtensions.\nWhen a node receives an RREQ, it first verifies the signature before cre-\nating or updating a reverse route to that host. Only if the signature is verified\nwill it store the route. If the RREQ was received with a Double Signature\nExtension, then the node will also store the signature for the RREP and the\nlifetime (which is the “reverse route lifetime” value) in the route entry. An\nintermediate node will reply to an RREQ with an RREP only if it fulfills the\nAODV’s requirements to do so and the node has the corresponding signa-\nture and old lifetime to put into the Signature and Old Lifetime fields of the\nRREP Double Signature Extension. Otherwise, it will rebroadcast the RREQ.\n" }, { "page_number": 197, "text": "186\n■\nSecurity in Wireless Mesh Networks\nWhen an RREQ is received by the destination itself, it will reply with an\nRREP only if it fulfills the AODV’s requirements to do so. This RREP will\nbe sent with an RREP Single Signature Extension.\nWhen a node receives an RREP, it first verifies the signature before\ncreating or updating a route (also called direct route) to that host. Only\nif the signature is verified, will it store the route with the signature of the\nRREP and the lifetime.\nBoth in the case of reverse and direct routes, routes are stored because\nthey meet the import authorization requirement. That is, the route infor-\nmation that is being authorized in the routing table is about the node that\nis sending the information. In the case of reverse routes, it is about the\noriginator of the RREQ (which is the node toward which the reverse route\npoints). In the case of direct routes, it is about the originator of the RREP\n(which is the node towards which the direct route points).\nIn this way, if either the originator of the RREQ or the originator of the\nRREP messages gives fake information in those messages, the only thing\nthat they might cause is that others will not be able to route packets to\nthem.\nUsing digital signatures prevents attack scenarios 1 and 3.\n5.9.3\nSecuring Error Messages\nConcerning RERR messages, someone could think that the right approach\nto secure them should be similar to the way the other AODV messages are\n(signing the non-mutable information and finding out a way to secure the\nmutable information). Nevertheless, RERR messages have a large amount of\nmutable information. In addition, it is not relevant which node started the\nRERR and which nodes are just forwarding it. The only relevant information\nis that a neighbor node is informing another node that it is not going to be\nable to route messages to certain destinations anymore.\nSAODV’s proposal is that every node (generating or forwarding an RERR\nmessage) will use digital signatures to sign the whole message and that any\nneighbor that receives it will verify the signature. In this way it can verify\nthat the sender of the RERR message is really the one that it claims to be.\nBecause destination sequence numbers are not signed by the correspond-\ning node, a node should never update any destination sequence number\nof its routing table based on an RERR message (this prevents a malicious\nnode from performing attack type 6). Implementing a mechanism that will\nallow the destination sequence numbers of an RERR message to be signed\nby their corresponding nodes would add too much overhead compared\nwith the advantage of the use of that information.\nAlthough nodes will not trust destination sequence numbers in an RERR\nmessage, they will use them to decide whether or not they should invalidate\na route. This does not give any extra advantage to a malicious node.\n" }, { "page_number": 198, "text": "Secure Routing in Wireless Mesh Networks\n■\n187\n5.9.4\nPersistence of Sequence Numbers\nThe attack type 7 was based on the fact that the originator of the RREQ\ncan set the sequence number of the destination. This should have not been\nspecified in AODV because it is not needed. In the case where everybody\nbehaves according to the protocol, the situation in which the originator of\na RREQ will put a destination sequence number bigger than the real one\nwill never happen, not even in the case that the destination of the RREQ\nhas rebooted. After rebooting, the node does not remember its sequence\nnumber anymore, but it waits long enough before being active, so that\nwhen it wakes up, nobody has stored its old sequence number anymore.\nTo avoid this attack, in the case that the destination sequence number in\nthe RREQ is bigger than the destination sequence number of the destination\nnode, the destination node will not take into account the value in the RREQ.\nInstead, it will realize that the originator of the RREQ is misbehaving and\nwill send the RREP with the right sequence number.\nIn addition, if one of the nodes has a way to store its sequence number\nevery time it modifies it, it might do so. Therefore, when it reboots it, will\nnot need to wait long enough so that everybody deletes routes toward it.\n5.10\nOpen Issues\nThe digital signature Digital signatureX (routing message) can be created\nonly by X. Thus, it serves as proof of validity of the information contained\nin the routing message. This prevents attack scenarios 1, 3, 4, and 6.\nThe hop authenticator reduces the ability of a malicious intermediate\nhop to mount the attack type 2 by arbitrarily modifying the hop count\nwithout detection. A node that is n hops away from T will know the nth\nelement in the hash chain (hn(x)), but it will not know any element that\ncomes before this because of the one-way property of h ( ). However, the\nmalicious node could still pass on the received authenticator and hop count\nwithout modifying it. Thus, the effectiveness of this approach is limited.\nIn addition, there is another type of attack that cannot be detected\nby SAODV: tunneling attacks. In that type of attack, two malicious nodes\nsimulate that they have a link between them (that is, they can send and\nreceive messages directly to each other). They achieve this by tunneling\nAODV messages between them (probably in an encrypted way). In this\nway they could achieve having certain traffic through them.\nNo security scheme has been able, so far, to detect this attack. Misbe-\nhaving detection schemes could, in principle, detect the so-called tunnel\nattacks. If the monitor sees a routing message with Hop Count = X + 1\nbeing sent by a node, but does not see a routing message with Hop Count\n= X being sent to the same node, then the node is either fabricating the\n" }, { "page_number": 199, "text": "188\n■\nSecurity in Wireless Mesh Networks\nrouting message or there is a tunnel. In either case, it is cause for raising\nthe alarm. Nevertheless, this kind of scheme has as main problems that\nthere is no way for any node to validate the authenticity of the misbehavior\nreports and there is the possibility of falsely detecting misbehavior nodes.\nTherefore, it is not a feasible solution so far.\nThe way the hop count is authenticated could be changed to a more\nsecure one. For instance, intermediate nodes forwarding the routing mes-\nsages could include the address of the next hop to which the message is\nforwarded and sign it [32]. Another possibility would be to use forward-\nsecure signature schemes [20]. A forward-secure signature scheme is like a\nhash chain, except that to prove that you are n hops away from the tar-\nget, you should sign the routing message with the key corresponding to\nthe nth link. Unlike in the hash chain case, the same signing key is not\ngiven to the next hop. Only the next signing key is given. This prevents\nthe attack based on the possibility that a malicious node does not increase\nthe hop count when it forwards a routing message. With this scheme, at\nany time the routing message has only one signature. The problem is, of\ncourse, efficiency. There are schemes where the message sizes are reason-\nably small, but signing and verification are quite expensive. Then there are\nother schemes where RSA signing could be used, but the public key needed\nto verify the signatures is size O(m), where m is the diameter of the network.\nAll those approaches would be very expensive (probably not even feasi-\nble), and still, they would not prevent tunneling attacks at all. Therefore,\nthe use of hash chains might be, so far, the option that deals best with the\ntrade-off between security and performance.\nThe use of sequence numbers should prevent most of the possible reply\nattacks. A node will discard a replied message if it has received an origi-\nnal message because the replied message will not be “fresh enough.” To\nmake the prevention of reply attacks stronger, a node could increase its\nsequence number in more situations than what AODV mandates (or even\nperiodically).\nPapadimitratos and Haas suggest in [27] that it is possible to mount an\nattack by maliciously modifying the IP header of the SAODV messages. This\nis not true because SAODV does not trust the contents of the IP header,\nand all the information that needs to operate is inside the AODV message\nand the SAODV extension.\n5.11\nAODV Message Formats\nFigures 5.5 through 5.8 show the structure of the AODV messages and\nindicate what the mutable fields of the messages are.\n" }, { "page_number": 200, "text": "Secure Routing in Wireless Mesh Networks\n■\n189\nOriginator sequence number\nOriginator IP address\nDestination IP address\nPREQ ID\nDestination sequence number\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nType\nJ R G\nReserved\nHop count\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\nFigure 5.5\nRoute request (RREQ) message format. Mutable fields: Hop count.\nLifetime\nOriginator IP address\nDestination IP address\nDestination sequence number\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nType\nR A\nReserved\nHop count\n6 7 8 9 0 1\nPrefix Sz\n2 3 4 5 6 7 8 9 0\n3\n1\n2\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\nFigure 5.6\nRoute reply (RREP) message format. Mutable fields: Hop count.\nAdditional unreachable destination sequence numbers (if needed)\nAdditional unreachable destination IP address (if needed)\nUnreachable destination IP address (1)\n Unreachable destination sequence number (1)\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nType\nN\nReserved\nDest count\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\nFigure 5.7\nRoute error (RERR) message format. Mutable fields: None.\n0\n1\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nType\nReserved\n1 2\n3 4 5\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+\nFigure 5.8\nRoute reply acknowledgment (RREP-ACK) message format. Mutable\nfields: None.\n" }, { "page_number": 201, "text": "190\n■\nSecurity in Wireless Mesh Networks\n5.12\nSecure AODV Extensions\nFigure 5.9 and Figure 5.10 and Table 5.2 show the format of the SAODV\nsignature extensions.\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nType\nHash\nHash function\nMax hop count\nTop hash\nLength\nSignature\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n. . .\n. . .\n. . .\n. . .\n. . .\n. . .\nFigure 5.9\nRREQ (single) signature extension.\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nType\nHash\nHash function\nMax hop count\nTop hash\nLength\nSignature\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n. . .\n. . .\n. . .\n. . .\n. . .\n. . .\nFigure 5.10\nRREP (single) signature extension.\nTable 5.2\nRREQ and RREP Signature Extension Fields\nField\nValue\nType\n64 in RREQ-SSE and 65 in RREP-SSE\nLength\nThe length of the type-specific data, not including the Type and\nLength fields of the extension.\nHash Function\nThe hash function used to compute the Hash and Top Hash fields.\nMax Hop Count\nThe Maximum Hop Count supported by the hop count\nauthentication.\nTop Hash\nThe top hash for the hop count authentication. This field has\nvariable length, but it must be 32-bits aligned.\nSignature\nThe signature of all the fields in the AODV packet that are\nbefore this field but the Hop Count field. This field has variable\nlength, but it must be 32-bits aligned.\nHash\nThe hash corresponding to the actual hop count. This field has\nvariable length, but it must be 32-bits aligned.\n" }, { "page_number": 202, "text": "Secure Routing in Wireless Mesh Networks\n■\n191\nFigure 5.11 and Table 5.3 show the format of the RREQ double signature\nextension.\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nHash\nTop hash\nReserved\nR A\nType\nLength\nHash function\nMax hop count\nPrefix Sz\nSignature for RREP\nSignature\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n. . .\n. . .\n. . .\n. . .\n. . .\n. . .\n. . .\n. . .\nFigure 5.11\nRREQ double signature extension.\nTable 5.3\nRREQ Double Signature Extension Fields\nField\nValue\nType\n66\nLength\nThe length of the type-specific data, not including the Type\nand Length fields of the extension.\nHash Function\nThe hash function used to compute the Hash and Top Hash\nfields.\nMax Hop Count\nThe Maximum Hop Count supported by the hop count\nauthentication.\nR\nRepair flag for the RREP.\nA\nAcknowledgment required flag for the RREP.\nReserved\nSent as 0; ignored on reception.\nPrefix Size\nThe prefix size field for the RREP.\nTop Hash\nThe top hash for the hop count authentication. This field\nhas variable length, but it must be 32-bits aligned.\nSignature\nThe signature of all the fields in the AODV packet that\nare before this field but the Hop Count field. This field has\nvariable length, but it must be 32-bits aligned.\nSignature\nThe signature that should be put into the Signature field of\nfor the RREP\nthe RREP Double Signature Extension when an intermediate\nnode (that has previously received this RREQ and created a\nreverse route) wants to generate an RREP for a route to the\nsource of this RREQ. This field has variable length, but it\nmust be 32-bits aligned. Both signatures are generated by\nthe requesting node.\nHash\nThe hash corresponding to the actual hop count. This field\nhas variable length, but it must be 32-bits aligned.\n" }, { "page_number": 203, "text": "192\n■\nSecurity in Wireless Mesh Networks\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nType\nHash\nTop hash\nHash function\nMax hop count\nLength\nSignature of the new lifetime\nOld lifetime\nSignature\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n. . .\n. . .\n. . .\n. . .\n. . .\n. . .\n. . .\n. . .\nFigure 5.12\nRREP double signature extension.\nTable 5.4\nRREP Double Signature Extension Fields\nField\nValue\nType\n67\nLength\nThe length of the type-specific data, not including\nthe Type and Length fields of the extension.\nHash Function\nThe hash function used to compute the Hash and\nTop Hash fields.\nMax Hop Count\nThe Maximum Hop Count supported by the hop\ncount authentication.\nTop Hash\nThe top hash for the hop count authentication. This\nfield has variable length, but it must be 32-bits\naligned.\nSignature\nThe signature of all the fields of the AODV packet\nthat are before this field but the Hop Count field,\nand with the Old Lifetime value instead of the\nLifetime. This signature is the one that was generated\nby the final destination. This field has variable\nlength, but it must be 32-bits aligned.\nOld Lifetime\nThe lifetime that was in the RREP generated by the\nfinal destination.\nSignature of the\nThe signature of the RREP with the actual lifetime\nNew Lifetime\n(the lifetime of the route in the intermediate node).\nThis signature is generated by the intermediate node.\nThis field has variable length, but it must be 32-bits\naligned.\nHash\nThe hash corresponding to the actual hop count.\nThis field has variable length, but it must be 32-bits\naligned.\n" }, { "page_number": 204, "text": "Secure Routing in Wireless Mesh Networks\n■\n193\nFigure 5.12 and Table 5.4 show the format of the RREP double signature\nextension.\nFinally, Figure 5.13 and Figure 5.14 and Table 5.5 show the format of\nthe RERR and RREP-ACK signature extensions.\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nSignature\nType\nLength\nReserved\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ + +\n‒\n‒\n. . .\n. . .\nFigure 5.13\nRERR signature extension.\n0\n1\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\nType\nReserved\n1 2\n3 4 5\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+\nFigure 5.14\nRREP-ACK signature extension.\nTable 5.5\nRERR and RREP-ACK Signature Extension Fields\nField\nValue\nType\n68 in RERR-SE and 69 in RREP-ACK-SE\nLength\nThe length of the type-specific data, not including the Type and Length\nfields of the extension.\nReserved\n(Only in RERR-SE). Sent as 0; ignored on reception.\nSignature\nThe signature of all the fields in the AODV packet that are before\nthis field. This field has variable length, but it must be 32-bits aligned.\nReferences\n[1]\nS. Asherson and A. Hutchison. Secure Routing for Wireless Mesh Networks.\nIn Proceedings of the Southern African Telecommunication Networks and\nApplications Conference (SATNAC) 2006, September 2006.\n" }, { "page_number": 205, "text": "194\n■\nSecurity in Wireless Mesh Networks\n[2]\nN. Asokan and P. Ginzboorg. Key agreement in ad-hoc networks. Computer\nCommunication Review, 23(17): 1627–1637, November 2000.\n[3]\nJ. Broch, D. A. Maltz, D. B. Johnson, Y. C. Hu, and J. Jetcheva. A Perfor-\nmance Comparison of Multi-hop Wireless Ad hoc Network Routing Proto-\ncols. In Proceedings of the 4th Annual International Conference on Mobile\nComputing and Networking, pp. 85–97, 1998.\n[4]\nJ. Chen, Y.-Z. Lee, D. Maniezzo, and M. Gerla. Performance Comparison of\nAODV and OFLSR in Wireless Mesh Networks. In Proceedings of the The\nFifth Annual Mediterranean Ad hoc Networking Workshop (Med-Hoc-Net\n2006), pp. 271–278, June 2006.\n[5]\nS. Cheung. An Efficient Message Authentication Scheme for Link State Rout-\ning. In 13th Annual Computer Security Applications Conference, pp. 90–98,\n1997.\n[6]\nT. Clausen, P. J. (Editors), C. Adjih, A. Laouiti, P. Minet, P. Muhlethaler,\nA. Qayyum, and L. Viennot. Optimized link state routing protocol (olsr).\nRFC 3626, October 2003. Network Working Group.\n[7]\nB. Dahill, B. N. Levine, E. Royer, and C. Shields. A Secure Routing Proto-\ncol for Ad hoc Networks. Technical report UM-CS-2001-037, University of\nMassachusetts, Departament of Computer Science, August 2001.\n[8]\nN. Ferguson and B. Schneier. A Cryptographic Evaluation of Ipsec. Tech-\nnical report, Counterpane Internet Security, February 1999.\n[9]\nM. Guerrero Zapata. Secure ad hoc on-demand distance vector routing.\nACM Mobile Computing and Communications Review (MC2R), 6(3): 106–\n107, July 2002.\n[10]\nM. Guerrero Zapata. Secure ad hoc on-demand distance vector (SAODV)\nrouting. First published in the IETF MANET Mailing List (October 8, 2001),\nAugust 2002. INTERNET-DRAFT — work in progress. draft-guerrero-manet-\nsaodv-00.txt.\n[11]\nM. Guerrero Zapata. Secure ad hoc on-demand distance vector (SAODV)\nrouting, Sept. 2006. INTERNET-DRAFT—work in progress. draft-guerrero-\nmanet-saodv-06.txt.\n[12]\nM. Guerrero Zapata and N. Asokan. Securing Ad hoc Routing Protocols. In\nProceedings of the 2002 ACM Workshop on Wireless Security (WiSe 2002),\npp. 1–10, September 2002.\n[13]\nR. Hauser, A. Przygienda, and G. Tsudik. Reducing the Cost of Security\nin Link State Routing. In Symposium on Network and Distributed Systems\nSecurity (NDSS ’97), pp. 93–99, San Diego, February 1997. Internet\nSociety.\n[14]\nY. C. Hu, D. Johnson, and A. Perrig. SEAD: Secure Efficient Distance Vector\nRouting for Mobile Wireless Ad hoc Networks. In 4th IEEE Workshop on\nMobile Computing Systems and Applications (WMCSA ’02), pp. 3–13, June\n2002.\n[15]\nY.-C. Hu and A. Perrig. A survey of secure wireless ad hoc routing. IEEE\nSecurity and Privacy, 2(3): 28–39, 2004.\n[16]\nY. C. Hu, A. Perrig, and D. Johnson. Ariadne: A Secure On-demand Routing\nProtocol for Ad hoc Networks. Technical report TR01-383, Rice University,\nDecember 2001.\n" }, { "page_number": 206, "text": "Secure Routing in Wireless Mesh Networks\n■\n195\n[17]\nD. B. Johnson et al. The dynamic source routing protocol (DSR) for mobile\nad hoc networks for IPv4. IETF Request for Comments, RFC4728, February\n2007.\n[18]\nC. Karlof and D. Wagner. Secure routing in wireless sensor networks:\nAttacks and countermeasures. Elsevier’s Ad Hoc Networks Journal, Spe-\ncial Issue on Sensor Network Applications and Protocols, 1(2–3): 293–315,\nSeptember 2003.\n[19]\nS. Kent and R. Atkinson. Security architecture for the internet protocol. IETF\nRequest for Comments, RFC 2401, November 1998.\n[20]\nH. Krawczyk. Simple Forward-Secure Signatures from Any Signature\nScheme. In ACM Conference on Computer and Communications Security,\npp. 108–115, 2000.\n[21]\nC. Madson and R. Glenn. The use of HMAC-MD5-96 within ESP and AH.\nInternet Request for Comments RFC 2403, November 1998.\n[22]\nC. Madson and R. Glenn. The use of HMAC-SHA-1-96 within ESP and AH.\nInternet Request for Comments RFC 2404, November 1998.\n[23]\nS. Marti, T. J. Giuli, K. Lai, and M. Baker. Mitigating Routing Misbehavior in\nMobile Ad hoc Networks. In Proceedings of the 6th Annual International\nConference on Mobile Computing and Networking, pp. 255–265, 2000.\n[24]\nP. Papadimitratos and Z. J. Haas. Secure Routing for Mobile Ad hoc Net-\nworks. SCS Communication Networks and Distributed Systems Modeling\nand Simulation Conference (CNDS 2002), January 2002.\n[25]\nC. E. Perkins, E. M. Belding-Royer, and S. R. Das. Ad hoc on-demand dis-\ntance vector (AODV) routing. Internet Request for Comments RFC 3561,\nNovember 2003.\n[26]\nC. E. Perkins and E. M. Royer. Ad hoc On-Demand Distance Vector Routing.\nIn Proceedings of the 2nd IEEE Workshop on Mobile Computing Systems and\nApplications, New Orleans, pp. 90–100, February 1999.\n[27]\nA. Perrig, R. Canetti, D. Song, and D. Tygar. Efficient and Secure Source\nAuthentication for Multicast. In Network and Distributed System Security\nSymposium (NDSS’01), February 2001.\n[28]\nA. Perrig, R. Szewczyk, V. Wen, D. E. Culler, and J. D. Tygar. SPINS: Se-\ncurity Protocols for Sensor Networks. In Proceedings of the 7th Annual\nInternational Conference on Mobile Computing and Networking, pp. 189–\n199, 2001.\n[29]\nS. Ramanathan and M. Steenstrup. A survey of routing techniques for mobile\ncommunications networks. Mobile Networks and Applications, 1(2): 89–104,\n1996.\n[30]\nE. M. Royer and C.-K. Toh. A review of current routing protocols for ad\nhoc mobile wireless networks. IEEE Personal Communications, pp. 46–55,\nApril 1999.\n[31]\nK. Sanzgiri, B. Dahill, B. Levine, and E. Belding-Royer. A secure routing\nprotocol for ad hoc networks. In International Conference on Network\nProtocols (ICNP), Paris, November 2002.\n[32]\nB. R. Smith, S. Murthy, and J. J. Garcia-Luna-Aceves. Securing distance-\nvector routing protocols. In Symposium on Network and Distributed Systems\nSecurity (NDSS ’97), pp. 85–92, San Diego, February 1997. Internet Society.\n" }, { "page_number": 207, "text": "196\n■\nSecurity in Wireless Mesh Networks\n[33]\nF. Stajano and R. Anderson. The Resurrecting Duckling: Security Issues for\nAd-hoc Wireless Networks. In Proceedings of the 7th International Work-\nshop on Security Protocols, number 1796 in Lecture Notes in Computer\nScience, pp. 172–194. Springer-Verlag, Berlin Germany, April 1999.\n[34]\nH. Yang, H. Luo, F. Ye, S. Lu, and L. Zhang. Security in mobile ad hoc\nnetworks: challenges and solutions. Wireless Communications, IEEE [see\nalso IEEE Personal Communications], 11(1): 38–47, 2004.\n[35]\nY. Yang, J. Wang, and R. Kravets. Designing Routing Metrics for Mesh Net-\nworks. In Proceedings of the First IEEE Workshop on Wireless Mesh Networks\n(WiMesh-2005), September 2005.\n[36]\nK. Zhang. Efficient Protocols for Signing Routing Messages. In Proceedings\nof the Symposium on Network and Distributed Systems Security (NDSS’98),\nJuly 2001.\n[37]\nL. Zhou and Z. J. Haas. Securing ad hoc networks. IEEE Network Magazine,\n13(6): 24–30, November/December 1999.\n" }, { "page_number": 208, "text": "Chapter 6\nHop Integrity in Wireless\nMesh Networks\nChin-Tser Huang\nContents\n6.1\nIntroduction ........................................................ 198\n6.2\nHop Integrity ....................................................... 201\n6.3\nInitial Authentication Protocol ..................................... 203\n6.4\nSecret Exchange Protocol .......................................... 208\n6.5\nIntegrity Check Protocol ........................................... 214\n6.5.1\nWeak Integrity Check Protocol ............................ 214\n6.5.2\nStrong Integrity Check Protocol ........................... 218\n6.6\nConclusion and Open Issues ...................................... 225\nReferences................................................................. 226\nMessage manipulation has become one of the major threats to the security\nof wireless mesh networks because of the open medium in such networks.\nAn adversary can launch a message insertion attack or a message replay\nattack such that the next mesh router that receives an inserted or replayed\nmessage will unwittingly forward it toward the destination. Even if a mes-\nsage from these attacks fails the authentication at the destination and gets\ndiscarded, it has already consumed the communication resources along\nthe forwarding path in the wireless mesh network. Repeated attempts of\nthese types of attack may result in a denial-of-service attack that may\nparalyze the network. To counter these attacks, it is necessary to provide\nmessage authentication and message integrity at every hop. In this chapter,\n197\n" }, { "page_number": 209, "text": "198\n■\nSecurity in Wireless Mesh Networks\nwe first address the need of sufficient and efficient authentication and\nintegrity checks at every hop by presenting several attack scenarios and\nexplaining possible constraints on wireless mesh routers. Then, we present\na novel protocol suite aimed to provide hop integrity for multi-hop wireless\nmesh networks. This protocol suite consists of three protocols: (1) an ini-\ntial authentication protocol for a joining mesh router to use a certificate to\nachieve mutual authentication and set up an initial shared secret with each\nof its adjacent mesh routers; (2) a secret exchange protocol used by two\nadjacent mesh routers to periodically update the secret they share for the\npurpose of computing message digests; and (3) an integrity check protocol\nused for computing and verifying message digests and sequence numbers.\nTogether, these three protocols can provide hop integrity for wireless mesh\nnetworks to counter message insertion attacks and message replay attacks.\nFurthermore, these three protocols are specified using a formal notation\ncalled Abstract Protocol Notation, and the correctness of these protocols is\nverified with state transition diagrams.\n6.1\nIntroduction\nWireless mesh networks [1–3] are networks consisting of mesh routers.\nSome of the mesh routers may be connected to the wired infrastructure\nof the Internet, but most of them are not. These ad hoc mesh routers\nare able to dynamically self-organize and self-configure, which is one of\nthe major advantages of wireless mesh networks. By forwarding packets\nvia mesh routers, wireless mesh networks provide communication paths\nto client nodes that are not within direct radio transmission range with\nanother client node or an Internet attachment point. As the popularity of\nwireless mesh networks grows, there are more and more attacks directed at\nwireless mesh networks and the security of them draws increased concern.\nIn particular, message manipulation has become one of the major threats\nto the security of wireless mesh networks because of the open medium in\nsuch networks. The threat of message manipulation can be realized by the\nfollowing two attacks:\n1.\nMessage insertion attack: An adversary impersonates a legitimate\nmesh router and inserts messages fabricated by itself. Alternatively,\nthe adversary can intercept a message in transit, arbitrarily mod-\nify the content of the message, and insert the modified message\ninto the network.\n2.\nMessage replay attack: An adversary makes copies of legitimate mes-\nsages intercepted between one pair of adjacent mesh routers and\nreplays them between the same pair or another pair of adjacent\nmesh routers in the same wireless mesh network, thanks to the\nmulti-hop nature of such network.\n" }, { "page_number": 210, "text": "Hop Integrity in Wireless Mesh Networks\n■\n199\nThe next mesh router that receives an inserted or replayed message will\nunwittingly forward it toward its ultimate destination if no appropriate pro-\ntection is provided. Even if a message originated from one of the above\ntwo attacks fails the authentication and integrity check mechanism (such as\nIPsec [4–6]) at the destination and gets discarded, it has already consumed\nthe communication resources along the forwarding path in the wireless\nmesh network. If no appropriate protection is provided, repeated attempts\nof these types of attack may result in a denial-of-service attack that may\nparalyze the wireless mesh network. To counter these attacks, it is neces-\nsary to provide message authentication and integrity check at every hop of\nthe network.\nIn this chapter, we apply the concept of hop integrity to address the\nabove problems. This chapter consists of two major components. First, we\naddress the need for sufficient and efficient authentication and integrity\ncheck at every hop by presenting several attack scenarios and introducing\nthe concept of hop integrity. Second, we present a novel protocol suite\naimed to provide hop integrity for multi-hop wireless mesh networks. This\nprotocol suite consists of three protocols. The first protocol is an initial\nauthentication protocol used for a joining mesh router to use a certificate\nissued by the certificate authority to achieve mutual authentication and set\nup an initial shared secret with each of its adjacent mesh routers. The sec-\nond protocol is a secret exchange protocol used by two adjacent mesh\nrouters to periodically update the secret they share for the purpose of com-\nputing message digests. The third protocol is an integrity check protocol\nused for computing and verifying message digests. In the integrity check\nprotocol, a soft sequence number is attached to each message as a fresh-\nness identifier. Together, these three protocols can provide hop integrity\nfor wireless mesh networks to counter message insertion attack and mes-\nsage replay attack. Furthermore, these three protocols are specified using a\nformal notation and the correctness of these protocols is verified with state\ntransition diagrams.\nThe protocols in this chapter are specified using a version of the Abstract\nProtocol Notation presented in [7]. We use this notation because it provides\na well-defined set of semantics that is suitable for a distributed environment\nand is not provided by programming languages like C/C++. In this notation,\neach process in a protocol is defined by a set of inputs, a set of variables, a\nset of parameters, and a set of actions. For example, in a protocol consisting\nof two processes x and y, process x can be defined as follows.\nprocess x\ninp\n⟨name of input⟩\n:\n⟨type of input⟩\n. . .\n⟨name of input⟩\n:\n⟨type of input⟩\n" }, { "page_number": 211, "text": "200\n■\nSecurity in Wireless Mesh Networks\nvar\n⟨name of variable⟩\n:\n⟨type of variable⟩\n. . .\n⟨name of variable⟩\n:\n⟨type of variable⟩\npar\n⟨name of parameter⟩:\n⟨type of parameter⟩\n. . .\n⟨name of parameter⟩:\n⟨type of parameter⟩\nbegin\n⟨action⟩\n⟨action⟩\n. . .\n⟨action⟩\nend\nThe inputs of process x have constant values that are assigned by an\nupper layer process and can be changed, if necessary, only by the assigning\nprocess. An input can be read, but not written, by the actions of process x.\nThe variables of process x can be read and updated by the actions of\nprocess x. A parameter has a finite number of values and its use will be\ndescribed next. Comments can be added anywhere in a process definition;\neach comment is placed between the two brackets { and }.\nEach ⟨action⟩of process x is of the form:\n⟨guard⟩→⟨statement⟩\nThe guard of an action of x is either a Boolean expression over the\nconstants and variables of x, a receive guard of the form rcv ⟨message⟩\nfrom y, or a time-out guard of the form time-out ⟨time expression⟩. The\n⟨time expression⟩refers to a time period because some action has executed\nlast and a Boolean expression that involves the constants and variables\nof the process. A parameterized action that refers to one parameter is a\nshorthand notation for a finite set of actions: each of them refers to a\ndifferent value in the domain of the parameter.\nExecuting an action consists of executing the statement of this action.\nExecuting the actions (of different processes) in a protocol proceeds ac-\ncording to the following three rules. First, an action is executed only when\nits guard is true. Second, the actions in a protocol are executed one at\na time. Third, an action whose guard is continuously true is eventually\nexecuted.\nThe ⟨statement⟩of an action of x is a sequence of ⟨skip⟩, ⟨assignment⟩,\n⟨send⟩, ⟨selection⟩, or ⟨iteration⟩statements of the following forms:\n" }, { "page_number": 212, "text": "Hop Integrity in Wireless Mesh Networks\n■\n201\n⟨skip⟩\n:\nskip\n⟨send⟩\n:\nsend ⟨message⟩to y\n⟨assignment⟩\n:\n⟨list of variables of x⟩:=\n⟨list of expressions⟩\n⟨selection⟩\n:\nif ⟨Boolean expression⟩→\n⟨statement⟩\n. . .\n⟨Boolean expression⟩→\n⟨statement⟩\nfi\n⟨iteration⟩\n:\ndo ⟨Boolean expression⟩→\n⟨statement⟩\nod\n6.2\nHop Integrity\nBefore we present the protocols, we introduce the concept of hop integrity\nbetween adjacent wireless mesh routers as discussed in [8–10]. Hop in-\ntegrity is fundamental to the three protocols in the hop integrity protocol\nsuite that are aimed to counter the aforementioned attacks and strengthen\nthe security of wireless mesh networks. The basic idea of hop integrity is\nstraightforward: whenever a mesh router p receives a message m from an\nadjacent mesh router q, p should be able to determine whether m was in-\ndeed sent by q or it was modified or replayed by an adversary that operates\nbetween p and q.\nNext, we discuss the requirements of hop integrity. A wireless mesh\nnetwork is said to provide hop integrity if and only if the following two\nconditions hold for every pair of adjacent mesh routers p and q in the\nnetwork:\n1.\nDetection of message modification: Whenever mesh router q re-\nceives a message m claimed to be transmitted from mesh router p,\nq can determine correctly whether message m was modified by an\nadversary after it was sent by p and before it was received by q.\n2.\nDetection of message replay: Whenever mesh router q receives a\nmessage m claimed to be transmitted from mesh router p, and de-\ntermines that message m was not modified, then q can determine\ncorrectly whether message m is another copy of a message that is\nreceived earlier by q.\nThe above two conditions infer receiving integrity, in which whenever a\nreceiver receives a message from a sender, the receiver can verify whether\nm was indeed sent by the sender or it was modified or replayed by an\n" }, { "page_number": 213, "text": "202\n■\nSecurity in Wireless Mesh Networks\nadversary that operates between the receiver and the sender. Note that the\nsender and the receiver referred to in our presentation of hop integrity are\none hop away from each other, i.e., a message transmitted by the sender\ncan be received directly by the receiver without the forwarding of other\nnodes.\nNext, we present the three protocols that are used to provide hop in-\ntegrity for wireless mesh networks. These protocols belong to two thin\nlayers, namely, the secret exchange layer and the integrity check layer,\nthat need to be added to the network layer of the protocol stack of each\nmesh router in a wireless mesh network. The function of the secret ex-\nchange layer is to allow adjacent mesh routers to periodically generate and\nexchange (and so share) new secrets. The exchanged secrets are made\navailable to the integrity check layer, which uses them to compute and\nverify the integrity check for every data message transmitted between the\nadjacent mesh routers.\nFigure 6.1 shows the protocol stacks in two adjacent mesh routers p\nand q. The secret exchange layer has two protocols: the initial authentica-\ntion protocol and the secret exchange protocol. The initial authentication\nprotocol consists of the two processes pa and qa, and the secret exchange\nprotocol consists of the two processes pe and qe in mesh routers p and q,\nrespectively. The integrity check layer has two protocols: the weak integrity\ncheck protocol and the strong integrity check protocol. The weak version\nconsists of the two processes pw and qw in mesh routers p and q, re-\nspectively. This version can detect message modification, but not message\nMAC and physical layer\nqw\nor \nqs\nMAC and physical layer\npw\nor \nps\nNetwork \nNetwork \nIntegrity \ncheck \nlayer \nSecret \nexchange \nlayer \nqa, qe\nTransport \nTransport \nApplication \nApplication \nMesh router q\nMesh router p\npa, pe\nshared \nsecret \nshared \nsecret \nFigure 6.1\nProtocol stack for hop integrity protocols.\n" }, { "page_number": 214, "text": "Hop Integrity in Wireless Mesh Networks\n■\n203\nreplay. The strong version of the integrity check layer consists of the two\nprocesses ps and qs in mesh routers p and q, respectively. This version\ncan detect both message modification and message replay.\nIn Section 6.3, we present the initial authentication protocol. In Section\n6.4, we present the light-weight secret exchange protocol. In Section 6.5,\nwe present the two versions of the integrity check protocol: weak version\nand strong version. The combination of these three protocols constitutes a\nprotocol suite that provides hop integrity to wireless mesh networks.\n6.3\nInitial Authentication Protocol\nBefore two adjacent mesh routers can forward messages to each other for\nthe first time, they have to use the initial authentication protocol to authen-\nticate each other. When a mesh router moves to a different location in the\nnetwork or is replaced by another mesh router, the initial authentication\nprotocol also needs to be executed. The initial authentication protocol is\ndesigned to achieve three things. First, it assures the two mesh routers that\nthey are communicating with a legitimate mesh router. Second, it allows the\ntwo mesh routers to exchange their certified public key. Third, it sets up\nthe initial shared secrets that will later be periodically updated by the secret\nexchange protocol. There are other upper layer protocols that provide au-\nthentication; for example, TLS [11] at the transport layer and Kerberos [12]\nat the application layer. However, those protocols provide end-to-end au-\nthentication and do not fit our needs well. In our case, we want to provide\nauthentication at the network layer for each pair of adjacent mesh routers\nthat are only one hop away.\nIn many authentication protocols, an online authentication server is\ncommonly used to provide authentication service for clients or other servers.\nExamples of this design include Kerberos [12] and RADIUS [13]. However,\nin the context of wireless mesh networks, initial authentication does not\noccur frequently because most mesh routers are relatively static. Therefore,\nwe choose to use certificates to achieve this purpose. A certificate is simply\nthe binding of a host’s identifier and a host’s public key, with an expiration\ntime specified, and is signed by a certificate authority using its private key.\nThe most common type of certificate is called X.509, whose format and de-\ntails can be found at [14,15]. If the recipient of a certificate belongs to the\nsame domain as the sender (namely, the owner) of the certificate, it should\nknow the public key of the certificate authority and can use the certificate\nauthority’s signature to verify whether it is a legitimate and valid certificate\nand whether to accept and use the public key contained in the certificate. In\ncase a certificate is stolen and spoofed by an adversary, a challenge-and-\nresponse scheme, as is used in the initial authentication protocol, can be\nused to counter this attack. (Note that a mesh router can renew its expiring\n" }, { "page_number": 215, "text": "204\n■\nSecurity in Wireless Mesh Networks\ncertificate with the certificate authority in an offline manner, but this is\nbeyond the scope of our discussion.)\nIn the initial authentication protocol, each mesh router has a process\nresponsible for executing the protocol. Before two adjacent mesh routers\nperform initial authentication, they undergo an association procedure to\nnegotiate necessary parameters for MAC layer and PHY layer. During the\nassociation procedure they also exchange the router identifier. The mesh\nrouter with a larger identifier will perform active initial authentication; we\ncall this mesh router p and its authentication process pa. The mesh router\nwith a smaller identifier will perform passive initial authentication; we call\nthis mesh router q and its authentication process qa. An authentication\nrequest message sent by the mesh router with a smaller identifier will simply\nbe dropped to avoid conflict.\nBecause the communication between mesh router p and mesh router\nq is bidirectional, two shared secrets, one for each direction, need to be\ngenerated and maintained. (How the two shared secrets are used will be\nexplained in the next section.) Processes pa and qa both have a public\nkey and a private key that they use to encrypt and decrypt the messages\nthat carry the new secrets between them. A public key has to be certified\nby the certificate authority in the form of a certificate, whereas a private\nkey is known only to its owner process. The public and private keys of\nprocess pa are named Bp and R p, respectively; similarly, the public and\nprivate keys of process qa are named Bq and Rq, respectively.\nThere are five steps in the initial authentication protocol. In the first\nstep, process pa sends a request message rqst(CERTp, e) to process qa,\nwhere CERTp is the certificate of mesh router p and e is the encryption of\nthe concatenation of p’s identifier and a time stamp. The identifier is used\nto verify that p is indeed the owner of the certificate, and the time stamp is\nused both as a freshness identifier to protect against message replay attacks\nand as a challenge to protect against certificate spoofing attacks. Process pa\nencrypts the identifier and time stamp using its private key R p to provide a\nsignature that this message is generated by pa and protect it from arbitrary\nmodification by an adversary.\nIn the second step, process qa receives the request message from pa,\ndecrypts p’s certificate to derive public key Bp, and uses Bp to decrypt the\nidentifier and the time stamp. Process qa verifies that p is the owner of the\ncertificate and that the certificate is still valid. If successful, qa will use a\nrandom function to generate a new shared secret sp, and qa sends a reply\nmessage rply(CERTq, d, e) to pa, where CERTq is the certificate of mesh\nrouter q, d is the encryption of the concatenation of q’s identifier and the\nsame time stamp which qa received from pa in the request message, and e\nis the shared secret sp encrypted using pa’s public key Bp. The same time\nstamp is used here as a response to the challenge. Field d is encrypted using\nqa’s private key Rq to provide a signature that this message is generated\n" }, { "page_number": 216, "text": "Hop Integrity in Wireless Mesh Networks\n■\n205\nby qa and protect it from arbitrary modification by an adversary. Field e is\nencrypted using p’s public key Bp to ensure that only pa can derive the\nshared secret sp.\nIn the third step, pa receives the reply message rply(c, d, e) from qa,\ndecrypts q’s certificate to derive public key Bq, and uses Bq to decrypt the\nidentifier and the time stamp. Process pa verifies that q is the owner of the\ncertificate and that the certificate is still valid. If successful, pa decrypts e\nusing its private key R p to derive the shared secret generated by qa, and\nuses a random function to generate a new shared secret sq. Then, pa sends\na first acknowledgment message ack(e) to qa, where e is the encryption of\nthe concatenation of the shared secret sp received from qa and the shared\nsecret sq generated by pa.\nIn the fourth step, qa receives the first acknowledgment message ack(e)\nfrom pa, and uses its private key Rq to decrypt e and verify that the first half\nof the result is equal to the shared secret sp it generated earlier. This ensures\nqa that pa has successfully received and installed sp. Then, qa derives the\nshared secret generated by pa from the second half of the result, uses pa’s\npublic key Bp to encrypt this value, and sends the encrypted result in a\nsecond acknowledgment to pa.\nIn the fifth step, pa receives the second acknowledgment message\nsack(e) from qa, and uses its private key R p to decrypt e and verify that\nthe result is equal to the shared secret sp it generated earlier. The success\nof the fifth step ensures pa that qa has successfully received and installed\nthe shared secret sq and concludes the initial authentication between pa\nand qa.\nIn addition, if the initial authentication between pa and qa has not com-\npleted for an extended period of time (for example four times of the round\ntrip time between pa and qa), then it is an indication that one of the above\nfive messages was lost, and pa times out to resend the rqst message to qa.\nProcess pa and process qa in the initial authentication protocol can be\ndefined as follows:\nprocess\npa\ninp\nBa\n: integer\n{public key of authentication authority}\nBp, R p\n: integer\n{public key and private key of p}\nCERT p\n: integer\n{certificate’s value = NCR(Ra, (Bp; IDp; expp))}\nIDp\n: integer\n{identifier of p}\ntr\n: integer\n{upper bound on round-trip time}\nvar\nts\n: integer\n{current value of p’s system clock}\nexp\n: integer\n{expiration time of q’s certificate}\nsp\n: integer\nsq\n: array [0 . . 1] of integer {initially sq[0] = sq[1] = 0}\nc, d, e\n: integer\n" }, { "page_number": 217, "text": "206\n■\nSecurity in Wireless Mesh Networks\nt, id\n: integer\nBq\n: integer {public key of q}\nIDq\n: integer {identifier of q}\nbegin\n(process pa and process qa have not performed initial authentication) →\nts := TMSTP;\ne := NCR(R p, (ts; IDp));\nsend rqst(CERT p, e) to qa\nrcv rply(c, d, e) from qa →\n(Bq, IDq, exp) := DCR(Ba, c);\n(t, id) := DCR(Bq, d);\nif t ̸= ts ∨id ̸= IDq ∨(current time > exp) →\n{authentication fails} skip\nt = ts ∧id = IDq ∧(current time ≤exp) →\n{authentication succeeds}\nsp := DCR(R p, e);\nsq[0] := any ;\nsq[1] := sq[0];\ne := NCR(Bq, (sp; sq[0]));\nsend ack(e) to qa\nfi\nrcv sack(e) from qa →\nd := DCR(R p, e);\nif d = sq[0] →\n{secret exchange succeeds} skip\nd ̸= sq[0] →\n{secret exchange fails} skip\nfi\ntimeout ((4*tr seconds passed since rqst message sent last) ∧\n(pa and qa have not completed initial authentication)) →\nts := TMSTP;\ne := NCR(R p, (ts; IDp));\nsend rqst(CERT p, e) to qa\nend\nprocess qa\ninp\nBa\n: integer\n{public key of authentication authority}\nBq, Rq\n: integer\n{public key and private key of q}\nCERT q\n: integer\n{certificate’s value = NCR(Ra, (Bq; IDq; expq))}\nIDq\n: integer\n{identifier of p}\ntr\n: integer\n{upper bound on round-trip time}\n" }, { "page_number": 218, "text": "Hop Integrity in Wireless Mesh Networks\n■\n207\nvar\nts\n: integer\n{time stamp received from p}\nexp\n: integer\n{expiration time of p’s certificate}\nsq\n: integer\nsp\n: array [0 . . 1] of integer {initially sp[0] = sp[1] = 0}\nc, d, e\n: integer\nid\n: integer\nBp\n: integer {public key of p}\nIDp\n: integer {identifier of p}\nbegin\nrcv rqst(d, e) from pa →\n(Bp, IDp, exp) := DCR(Ba, d);\n(ts, id) := DCR(Bp, e);\nif id ̸= IDp ∨(current time > exp) →\n{authentication fails} skip\nid = IDp ∧(current time ≤exp) →\n{authentication succeeds}\nd := NCR(Rq, (ts; IDq));\nsp[0] := any ;\nsp[1] := sp[0];\ne := NCR(Bp, sp[0]);\nsend rply(Certq, d, e) to pa\nfi\nrcv ack(e) from pa →\n(c, d) := DCR(Rq, e);\nif c ̸= sp[0] →\n{secret exchange fails} skip\nc = sp[0] →\n{secret exchange succeeds}\nsq := d;\ne := NCR(Bp, sq);\nsend sack(e) to pa\nfi\nend\nProcesses pa and qa use three functions, namely, TMSTP, NCR, and\nDCR. Function TMSTP takes no arguments, and when invoked, it returns\na time stamp that is according to the system clock and is larger than any\ntime stamp generated by the same process in the past. In other words, the\ntime stamps generated by the same process are monotonic. Function NCR\nis an encryption function that takes two arguments, a key and a data item,\nand returns the encryption of the data item using the key. For example,\n" }, { "page_number": 219, "text": "208\n■\nSecurity in Wireless Mesh Networks\nexecution of the statement\ne := NCR(R p, (ts; IDp))\ncauses the concatenation of ts and IDp to be encrypted using the private\nkey R p, and the result to be stored in variable e. Function DCR is a decryp-\ntion function that takes two arguments, a key and an encrypted data item,\nand returns the decryption of the data item using the key. For example,\nexecution of the statement\nd := DCR(R p, e)\ncauses the (encrypted) data item e to be decrypted using the private key\nR p, and the result to be stored in variable d. As another example, consider\nthe statement\n(d, e) := DCR(R p, e)\nThis statement indicates that the value of e is the encryption of the\nconcatenation of two values (v0; v1) using key R p. Thus, executing this\nstatement causes e to be decrypted using key R p, and the resulting first\nvalue v0 to be stored in variable d, and the resulting second value v1 to be\nstored in variable e.\nNote in particular that in the specification of the initial authentication\nprotocol, process pa has the following variable declaration:\nvar\nsp\n:\ninteger\nsq\n:\narray [0 . . 1] of integer {initially sq[0] = sq[1] = 0}\nIn process qa, the array sp is defined in a similar way. Array sq in\nprocess pa and array sp in process qa will be used in the secret exchange\nprotocol and will be explained next.\n6.4\nSecret Exchange Protocol\nIn the secret exchange protocol, processes pe and qe maintain two shared\nsecrets sp and sq. Secret sp is used by mesh router p to compute the\nintegrity check for each data message sent by p to mesh router q, and it\nis also used by mesh router q to verify the integrity check for each data\nmessage received by q from mesh router p. Similarly, secret sq is used by q\nto compute the integrity checks for data messages sent to p, and it is used\nby p to verify the integrity checks for data messages received from q.\n" }, { "page_number": 220, "text": "Hop Integrity in Wireless Mesh Networks\n■\n209\nRecall that the two initial shared secrets sp and sq have been set up by\nthe initial authentication protocol. However, any shared secret grows more\nvulnerable to statistical attacks as the usage increases. As part of maintaining\nthe two secrets sp and sq, processes pe and qe need to change these secrets\nperiodically, say every te hours, for some chosen value te. Process pe is\nto initiate the change of secret sq, and process qe is to initiate the change\nof secret sp. Processes pe and qe both have a public key and a private\nkey that they use to encrypt and decrypt the messages that carry the new\nsecrets between pe and qe. These keys assume the same names and values\nas defined in the initial authentication protocol.\nFor process pe to change secret sq, the following four steps need to be\nperformed. First, pe generates a new sq, and encrypts the concatenation of\nthe old sq and the new sq using qe’s public key Bq, and sends the result\nin a rqst message to qe. Second, when qe receives the rqst message, it\ndecrypts the message contents using its private key Rq and obtains the old\nsq and the new sq. Then, qe checks that its current sq equals the old sq\nfrom the rqst message, and installs the new sq as its current sq, and sends\na rply message containing the encryption of the new sq using pe’s public\nkey Bp. Third, pe waits until it receives a rply message from qe contain-\ning the new sq encrypted using Bp. Receiving this rply message indicates\nthat qe has received the rqst message and has accepted the new sq. Fourth,\nif pe sends the rqst message to qe, but does not receive the rply message\nfrom qe for some tr seconds, indicating that either the rqst message or\nthe rply message was lost before it was received, then pe resends the rqst\nmessage to qe. Thus tr is an upper bound on the round-trip time between\npe and qe.\nNote that the old secret (along with the new secret) is included in each\nrqst message and the new secret is included in each rply message to ensure\nthat if an adversary modifies or replays rqst or rply messages, then each of\nthese messages is detected and discarded by its receiving process (whether\npe or qe).\nProcess pe has two variables sp and sq declared as follows:\nvar\nsp\n:\ninteger\nsq\n:\narray [0 . . 1] of integer\nSimilarly, process qe has an integer variable sq and an array variable sp.\nIn process pe, variable sp is used for storing the secret sp, variable\nsq[0] is used for storing the old sq, and variable sq[1] is used for storing the\nnew sq. The assertion sq[0] ̸= sq[1] indicates that process pe has generated\nand sent the new secret sq, and that qe may not have received it yet. The\nassertion sq[0] = sq[1] indicates that qe has already received and accepted\n" }, { "page_number": 221, "text": "210\n■\nSecurity in Wireless Mesh Networks\nthe new secret sq. Initially,\nsq[0] in pe = sq[1] in pe = sq in qe,\nand\nsp[0] in qe = sp[1] in qe = sp in pe.\nProcess pe can be defined as follows. (Process qe can be defined\nin the same way except that each occurrence of R p in pe is replaced by\nan occurrence of Rq in qe, each occurrence of Bq in pe is replaced by an\noccurrence of Bp in qe, each occurrence of sp in pe is replaced by an oc-\ncurrence of sq in qe, and each occurrence of sq[0] or sq[1] in pe is replaced\nby an occurrence of sp[0] or sp[1], respectively, in qe.)\nprocess\npe\ninp\nR p\n: integer\n{private key of p}\nBq\n: integer\n{public key of q}\nte\n: integer\n{time between secret exchanges}\ntr\n: integer\n{upper bound on round-trip time}\nvar\nsp\n: integer\nsq\n: array [0 . . 1] of integer {initially sq[0] = sq[1] = sq in qe}\nd, e\n: integer\nbegin\ntimeout (sq[0] = sq[1] ∧(te hours passed since rqst message sent last))→\nsq[1] := NEWSCR;\ne := NCR(Bq, (sq[0]; sq[1]));\nsend rqst(e) to qe\nrcv rqst(e) from qe →\n(d, e) := DCR(R p, e);\nif sp = d ∨sp = e →\nsp := e;\ne := NCR(Bq, sp);\nsend rply(e) to qe\nsp ̸= d ∧sp ̸= e →\n{detect adversary}\nskip\nfi\nrcv rply(e) from qe →\nd := DCR(R p, e);\nif sq[1] = d →\nsq[0] := sq[1]\n" }, { "page_number": 222, "text": "Hop Integrity in Wireless Mesh Networks\n■\n211\nsq[1] ̸= d →\n{detect adversary}\nskip\nfi\ntimeout (sq[0] ̸= sq[1] ∧(tr seconds passed since rqst\nmessage sent last)) →e := NCR(Bq, (sq[0]; sq[1]));\nsend rqst(e) to qe\nend\nThe four actions of process pe use three functions, namely, NEWSCR,\nNCR, and DCR. Function NEWSCR takes no arguments, and when invoked,\nit returns a fresh secret that is different from any secret that was returned in\nthe past. Functions NCR and DCR have been described in the last section.\nTo verify the correctness of the secret exchange protocol, we can use\nthe state transition diagram of this protocol in Figure 6.2. This diagram has\nsix nodes that represent all possible reachable states of the protocol. Every\ntransition in the diagram stands for either a legitimate action (of process pe\nor process qe), or an illegitimate action of the adversary.\nInitially, the protocol starts at a state S.0, where the two channels be-\ntween processes pe and qe are empty and the values of variables sq[0] and\nsq[1] in pe and variable sq in qe are the same. This state can be defined\nby the following predicate:\nS.0 :\nch.pe.qe =<> ∧ch.qe.pe =<> ∧\nsq[0] in pe = sq[1] in pe = sq in qe\nAt state S.0, exactly one action, namely, the first time-out action in pro-\ncess pe, is enabled for execution. Executing this action at state S.0 leads\nthe protocol to state S.1 defined as follows:\nS.1 :\nch.pe.qe =< rqst(e) > ∧ch.qe.pe =<> ∧\ne = NCR(Bq, (sq[0]; sq[1])) ∧\nsq[0] in pe ̸= sq[1] in pe ∧sq[0] in pe = sq in qe\nAt state S.1, exactly one legitimate action, namely, the receive action\n(that receives a rqst message) in process qe, is enabled for execution. Ex-\necuting this action at state S.1 leads the protocol to state S.2 defined as\nfollows:\nS.2 :\nch.pe.qe =<> ∧ch.qe.pe =< rply (e) > ∧\ne = NCR(Bp, sq) ∧\nsq[0] in pe ̸= sq[1] in pe ∧sq[1] in pe = sq in qe\n" }, { "page_number": 223, "text": "212\n■\nSecurity in Wireless Mesh Networks\nM:rqst\nL:rqst\ntimeout &\nS:rqst\nR:rqst & S:rply\nL:rply\nR:rqst\nR:rply\nR:rply\nM:rply\nP:rply\nP:rqst\nS.1\nS.0\nS.2\nM.1\nL.0\nM.2\ntimeout & S:rqst\nS.0 =\nS.1 =\nS.2 =\nM.1 =\nM.2 =\nL.0 =\nch.pe.qe = < >\nch.qe.pe = < >\nch.qe.pe = < >\nch.qe.pe = < >\nch.pe.qe = < >\nch.pe.qe = < >\nch.qe.pe = < >\nch.pe.qe = < >\nch.pe.qe = < rqst(e)>\nch.pe.qe = < rqst(e)>\nch.qe.pe = < rply(e)>\nch.qe.pe = < rply(e)>\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∧\n∨\n∨\n∨\n∧\n∧\n∧\n∧sq[0] in pe = sq[1] in pe = sq in qe \nsq[0] in pe ≠ sq[1] in pe\nsq[0] in pe ≠ sq[1] in pe\nsq[0] in pe ≠ sq[1] in pe\nsq[0] in pe ≠ sq[1] in pe\ne ≠ NCR(Bp, sq)\ne ≠ NCR(Bq, (sq[0]; sq[1]))\ne = NCR(Bp, sq)\n(sq[0] in pe = sq in qe\n(sq[0] in pe = sq in qe\nsq[1] in pe = sq in qe)\nsq[1] in pe = sq in qe)\nsq[0] in pe = sq in qe\nsq[1] in pe = sq in qe\nsq[1] in pe = sq in qe)\n(sq[0] in pe = sq in qe\nsq[0] in pe ≠ sq[1] in pe\ne = NCR(Bq, (sq[0]; sq[1]))\nFigure 6.2\nState transition diagram of the secret exchange protocol.\nAt state S.2, exactly one legitimate action, namely, the receive action\n(that receives a rply message) in process pe, is enabled for execution.\nExecuting this action at state S.2 leads the protocol back to state S.0 defined\nabove. States S.0, S.1, and S.2 are called good states because the transitions\nbetween these states consist of executing the legitimate actions of the two\nprocesses. The sequence of transitions from state S.0 to state S.1, to state\nS.2, and back to state S.0 constitutes the good cycle of the protocol. If only\nlegitimate actions of processes pe and qe are executed, the protocol will\nstay in this good cycle indefinitely. Next, we discuss the bad effects caused\n" }, { "page_number": 224, "text": "Hop Integrity in Wireless Mesh Networks\n■\n213\nby the actions of an adversary, and how the protocol can recover from\nthese effects.\nFirst, the adversary can execute a message loss action at state S.1 or\nS.2. If the adversary executes a message loss action at state S.1 or S.2, the\nnetwork moves to a state L.0 defined as follows:\nL.0 :\nch.pe.qe =<> ∧ch.qe.pe =<> ∧\nsq[0] in pe ̸= sq[1] in pe ∧\n(sq[0] in pe = sq in qe ∨sq[1] in pe = sq in qe)\nAt state L.0, only the second time-out action in pe is enabled for exe-\ncution, and executing this action leads the network back to state S.1.\nSecond, the adversary can execute a message modification action at\nstate S.1 or S.2. If the adversary executes a message modification action at\nstate S.1, the network moves to state M.1 defined as follows:\nM.1 :\nch.pe.qe =< rqst(e) > ∧ch.qe.pe =<> ∧\ne ̸= NCR(Bq, (sq[0]; sq[1])) ∧\nsq[0] in pe ̸= sq[1] in pe ∧\n(sq[0] in pe = sq in qe ∨sq[1] in pe = sq in qe)\nIf the adversary executes a message modification action at state S.2, the\nnetwork moves to state M.2 defined as follows:\nM.2 :\nch.pe.qe =<> ∧ch.qe.pe =< rply(e) > ∧\ne ̸= NCR(Bp, sq) ∧\nsq[0] in pe ̸= sq[1] in pe ∧\n(sq[0] in pe = sq in qe ∨sq[1] in pe = sq in qe)\nIn either case, the protocol moves next to state L.0 and eventually returns\nto state S.1.\nThird, the adversary can execute a message replay action at state S.1\nor S.2. If the adversary executes a message replay action at state S.1, the\nnetwork moves to state M.1. If the adversary executes a message replay\naction at state S.2, the network moves to state M.2. As shown above, the\nprotocol eventually returns to state S.1.\nFrom the state transition diagram in Figure 6.2, it is clear that each\nillegitimate action by the adversary will eventually lead the network back\nto state S.1, which is a good state. Once the network is in a good state, the\n" }, { "page_number": 225, "text": "214\n■\nSecurity in Wireless Mesh Networks\nnetwork can progress in the good cycle. Hence the following two theorems\nabout secret exchange protocol are proved:\nTheorem 1\nIn the absence of an adversary, a network that executes the secret exchange\nprotocol will follow the good cycle, consisting of the transitions from state S.0 to\nstate S.1, from state S.1 to state S.2, and from state S.2 to state S.0, and will stay\nin this good cycle indefinitely.\nTheorem 2\nIn the presence of an adversary, a network that executes the secret exchange\nprotocol will converge to the good cycle in a finite number of steps after the ad-\nversary finishes executing the message loss, message modification, and message\nreplay actions.\n6.5\nIntegrity Check Protocol\nThis section introduces the integrity check protocol, starting with a weak\nversion of the protocol, which detects message insertion only, and moving\non to a strong version of the protocol, which detects both message insertion\nand message replay.\n6.5.1\nWeak Integrity Check Protocol\nThe main idea of the weak integrity check protocol is simple. Consider the\ncase where a data(t) message, with t being the message text, is generated at\na source src, then transmitted through a sequence of adjacent mesh routers\nr.1, r.2, . . . , r.n to a destination dst. When data(t) reaches the first mesh\nrouter r.1, r.1 computes a digest d for the message as follows:\nd := MD(t; scr)\nwhere MD is the message digest function, (t; scr) is the concatenation of\nthe message text t and the shared secret scr between r.1 and r.2 (provided\nby the secret exchange protocol in r.1). Note that MD can be any common\nmessage digest function, such as MD5 [16], SHA [17], or HMAC [18]. Then, r.1\nadds d to the message before transmitting the resulting data(t, d) message\nto mesh router r.2.\nWhen r.2 receives the data(t, d) message, it computes the message di-\ngest using the secret shared between r.1 and r.2 (provided by the secret\nexchange process in r.2), and checks whether the result equals d. If they are\nunequal, then r.2 concludes that the received message has been modified,\ndiscards it, and reports an adversary. If they are equal, then r.2 concludes\nthat the received message has not been modified and proceeds to prepare\n" }, { "page_number": 226, "text": "Hop Integrity in Wireless Mesh Networks\n■\n215\nthe message for transmission to the next mesh router r.3. Preparing the\nmessage for transmission to r.3 consists of computing d using the shared\nsecret between r.2 and r.3 and storing the result in field d of the data(t, d)\nmessage. When the last mesh router r.n receives the data(t, d) message,\nit computes the message digest using the shared secret between r.(n −1)\nand r.n and checks whether the result equals d. If they are unequal, r.n\ndiscards the message and reports an adversary. Otherwise, r.n sends the\ndata(t) message to its destination dst.\nNote that this protocol detects and discards every modified message.\nMore importantly, it also determines the location where each message mod-\nification has occurred.\nProcess pw in the weak integrity protocol has two constants sp and\nsq that pw reads, but never updates. These two constants in process pw\nare also variables in process pe, and pe updates them periodically, as dis-\ncussed in the previous section. Process pw can be defined as follows.\n(Process qw is defined in the same way except that each occurrence of\np, q, pw, qw, sp, and sq is replaced by an occurrence of q, p, qw, pw, sq,\nand sp, respectively.)\nprocess\npw\ninp\nsp\n:\ninteger\nsq\n:\narray [0 . . 1] of integer\nvar\nt, d\n:\ninteger\nbegin\nrcv data(t, d) from qw →\nif MD(t; sq[0]) = d ∨MD(t; sq[1]) = d →\n{accept message}\nRTMSG\nMD(t; sq[0]) ̸= d ∧MD(t; sq[1]) ̸= d →\n{report an adversary}\nskip\nfi\ntrue →\n{p receives data(t, d) from mesh router other than q}\n{and checks that its message digest is correct}\nRTMSG\ntrue →\n{either p receives data(t) from an adjacent host or}\n{p generates the text t for the next data message}\nRTMSG\nend\n" }, { "page_number": 227, "text": "216\n■\nSecurity in Wireless Mesh Networks\nIn the first action of process pw, if pw receives a data(t, d) message from\nqw while sq[0] ̸= sq[1], then pw cannot determine beforehand whether qw\ncomputed d using sq[0] or using sq[1]. In this case, pw needs to compute\ntwo message digests using both sq[0] and sq[1], respectively, and compare\nthe two digests with d. If either digest equals d, then pw accepts the mes-\nsage. Otherwise, pw discards the message and reports the detection of an\nadversary.\nThe three actions of process pw use two functions named MD and\nNXT and one statement named RTMSG. Function MD takes one argument,\nnamely, the concatenation of the text of a message and the appropri-\nate secret, and computes a digest for that argument. Function NXT takes\none argument, namely, the text of a message (which we assume includes\nthe message header), and determines the next mesh router to which the\nmessage should be forwarded. Statement RTMSG is defined as follows:\nif NXT(t) = p →\n{accept message}\nskip\nNXT(t) = q →\nd := MD(t; sp);\nsend data(t, d) to qw\nNXT(t) ̸= p ∧NXT(t) ̸= q →\n{compute d as the message digest of the concatenation of t and\nthe secret}\n{for sending data to NXT(t); forward data(t, d) to mesh router NXT(t)}\nskip\nfi\nTo verify the correctness of the weak integrity check protocol, we can\nuse the state transition diagram of this protocol in Figure 6.3, which con-\nsiders the channel from process qw to process pw. (The channel from pw\nto qw and the channels from pw to any other weak integrity process in an\nadjacent mesh router of p can be verified in the same way.) This diagram\nhas two nodes that represent all possible reachable states of the proto-\ncol. Every transition in the diagram stands for either a legitimate action (of\nprocess pw or process qw), or an illegitimate action of the adversary.\nNote that because the weak integrity check protocol operates below\nthe secret exchange protocol in the protocol stack, we can assert that\n(sq in qw = sq[0] in pw ∨sq in qw = sq[1] in pw) is an invariant in ev-\nery state of the weak integrity protocol. We denote this invariant as I in the\nspecification in Figure 6.3. Also note that the notation Head(data(t, d)) in\n" }, { "page_number": 228, "text": "Hop Integrity in Wireless Mesh Networks\n■\n217\nsq in qw = sq[0] in pw\nsq in qw = sq[1] in pw\nI\n=\n=\n=\nwhere\nM.0\nT.0\nI\nI\nS:data\nS:data\nR:data & Accept\nR:data &\nDiscard\nM.0\nT.0\nL:data\nM:data\nd = MD(t; sq))\n(∀data(t, d) message in ch.qw.pw,\n(∀data(t, d) message in ch.qw.pw,\n( Head(data(t, d))\n( Head(data(t, d))\nd = MD(t; sq))\nd ≠ MD(t; sq)))\n∧\n∧\n∧\n∨\n⇒\n⇒\nFigure 6.3\nState transition diagram of the weak integrity check protocol.\nthe specification in Figure 6.3 is a predicate whose value is true if data(t, d)\nis the head message of the specified channel.\nInitially, the protocol starts at state T.0. At state T.0, two legitimate\nactions, namely, the send action in qw that sends a data message and the re-\nceive action in pw that receives a data message, can be executed. Executing\neither one of the two actions at state T.0 keeps the protocol in state T.0.\nState T.0 is the only good state in the weak integrity protocol. The\nsequence of the transitions from state T.0 to state T.0 constitutes the good\ncycle of the protocol. If only legitimate actions of processes pw and qw\nare executed, the protocol will stay in this good cycle indefinitely. Next,\nwe discuss the bad effects caused by the actions of an adversary, and how\nthe protocol can recover from these effects.\nFirst, the adversary can execute a message loss action at state T.0. In\nthis case, the predicate that for every data message data(t, d) in the channel\nfrom qw to pw, d = MD(t; sq), still holds. Therefore, the protocol stays at\nstate T.0.\nSecond, the adversary can execute a message modification action at\nstate T.0. In this case, the protocol moves to state M.0. The receive and dis-\ncard action executed by pw at state M.0 leads the protocol back to state T.0.\nFrom the state transition diagram, it is clear that each illegitimate action by\n" }, { "page_number": 229, "text": "218\n■\nSecurity in Wireless Mesh Networks\nthe adversary will eventually lead the protocol back to T.0, which is a good\nstate. Once the protocol is in a good state, the protocol can progress in the\ngood cycle. Hence the following two theorems about the weak integrity\ncheck protocol are proved:\nTheorem 3\nIn the absence of an adversary, a network that executes the weak integrity check\nprotocol follows the good cycle, consisting of the single transition from state T.0\nto state T.0, and will stay in this good cycle indefinitely.\nTheorem 4\nIn the presence of an adversary, a network that executes the weak integrity\ncheck protocol will converge to the good cycle in a finite number of steps after\nthe adversary finishes executing the message loss and message modification\nactions.\nHowever, the weak integrity check protocol, while being able to detect\nand discard all modified messages, cannot detect some replayed messages.\nThe next section introduces the strong integrity protocol that is capable of\ndetecting and discarding all modified and replayed messages.\n6.5.2\nStrong Integrity Check Protocol\nThe weak hop integrity protocol can detect message modification, but not\nmessage replay. This section discusses how to strengthen this protocol to\nmake it detect message replay as well. The strong hop integrity protocol\nis presented in two steps: (1) using “soft sequence numbers” to detect\nand discard replayed data messages, and (2) integrating this soft sequence\nnumber protocol into the weak integrity check protocol to construct the\nstrong integrity check protocol.\nBefore introducing the soft sequence number protocol, a simple proto-\ncol is used to illustrate the need for sequence numbers in detecting message\nreplay. Consider a protocol that consists of two processes u and v executing\non two adjacent mesh routers. Process u continuously sends data messages\nto process v. Because process u and process v are only one hop away, the\ndata messages sent by u will be received by v in the same order they were\nsent. Assume that there is an adversary that attempts to disrupt the commu-\nnication between u and v by inserting (i.e., replaying) old messages in the\nmessage stream from u to v. To overcome this adversary, process u attaches\nan integer sequence number s to every data message sent to process v. To\nkeep track of the sequence numbers, process u maintains a variable nxt\nthat stores the sequence number of the next data message to be sent by u\nand process v maintains a variable exp that stores the sequence number of\nthe expected data message to be received by v. (Note that a single variable\n" }, { "page_number": 230, "text": "Hop Integrity in Wireless Mesh Networks\n■\n219\nexp at process v is sufficient because there is no reorder.) This is called a\n“hard sequence number protocol,” because process u always remembers\nthe next sequence number to be sent, and process v always remembers\nthe next sequence number it expects to receive.\nTo send the next data(s) message, process u assigns s the current value\nof variable nxt, then increments nxt by one. When process v receives\na data(s) message, v compares its variable exp with s. If exp ≤s, then\nv accepts the received data(s) message and assigns exp the value s + 1;\notherwise, v discards the data(s) message. Processes u and v of this protocol\ncan be specified as follows:\nprocess u\nvar\nnxt\n: integer\n{sequence number of next sent message}\nbegin\ntrue →\nsend data(nxt) to v;\nnxt := nxt + 1\nend\nprocess v\nvar\ns\n: integer\n{sequence number of received message}\nexp\n: integer\n{sequence number expected next}\nbegin\nrcv data(s) from u →\nif s < exp →\n{reject message; report an adversary}\nskip\nexp ≤s →\n{accept message}\nexp := s + 1\nfi\nend\nCorrectness of this protocol is based on the observation that the pred-\nicate exp ≤nxt holds at each (reachable) state of the protocol. However,\nif due to some fault (for example, an accidental resetting of the values of\nvariable nxt) the value of exp becomes larger than value of nxt, then all\nthe data messages that u sends from this point and until the value of nxt\nbecomes equal to the value of exp will be wrongly discarded by v. Next\nis a description of how to modify this protocol such that the number of\n" }, { "page_number": 231, "text": "220\n■\nSecurity in Wireless Mesh Networks\nmessages, which can be wrongly discarded when the synchronization be-\ntween u and v is lost due to some fault, is at most N, for some chosen\ninteger N that is larger than one.\nThe modification consists of adding to process v two variables c and\ncmax, whose values are in the range 0..N-1. When process v receives a\ndata(s) message, v compares the values of c and cmax. If c ̸= cmax, then\nprocess v increments c by one (mod N) and proceeds as before, namely,\neither accepts the data(s) message if exp ≤s or discards the message if\nexp > s. Otherwise, if c = cmax, then v accepts the message, assigns c the\nvalue 0, and assigns cmax a random integer in the range 0..N-1. We call\nthis modified protocol “soft sequence number protocol” because process v\nat some instants “forgets” the sequence number it expects to receive next,\nand accepts the next received sequence number without question.\nThere are two considerations behind this modification. First, it guar-\nantees that process v never discards more than N data messages when\nthe synchronization between u and v is lost due to some fault. Second, it\nensures that the adversary cannot predict the instant when process v is will-\ning to accept any received data message, and so cannot exploit any such\npredictions by sending replayed data messages at the predicted instant.\nFormally, processes u and v in this protocol can be defined as follows:\nprocess u\nvar\nnxt\n: integer\n{sequence number of next sent message}\nbegin\ntrue →\nsend data(nxt) to v;\nnxt := nxt + 1\nend\nprocess v\ninp\nN\n: integer\nvar\ns\n: integer\n{sequence number of received message}\nexp\n: integer\n{sequence number expected next}\nc, cmax\n: 0 . . N-1\nbegin\nrcv data(s) from u →\nif s < exp ∧c ̸= cmax →\n{reject message; report an adversary}\nc := (c + 1)modN\nexp ≤s ∨c = cmax →\n" }, { "page_number": 232, "text": "Hop Integrity in Wireless Mesh Networks\n■\n221\n{accept message}\nexp := s + 1\nif c ̸= cmax →\nc := (c + 1)modN\nc = cmax →\nc := 0;\ncmax := RANDOM(0, N −1)\nfi\nfi\nend\nProcesses u and v of the soft sequence number protocol presented\nabove can be combined with process pw of the weak integrity check\nprotocol to construct process ps of the strong integrity check protocol.\nA main difference between processes pw and ps is that pw exchanges\nmessages of the form data(t, d), whereas ps exchanges messages of the\nform data(s, t, d), where s is the message sequence number computed ac-\ncording to the soft sequence number protocol, t is the message text, and\nd is the message digest computed over the concatenation (s; t; scr) of s, t,\nand the shared secret scr. Process ps in the strong integrity check protocol\ncan be defined as follows. (Process qs can be defined in the same way.)\nprocess\npw\ninp\nsp\n:\ninteger\nsq\n:\narray [0 . . 1] of integer\nN\n:\ninteger\nvar\ns, t, d\n:\ninteger\nexp, nxt\n:\ninteger\nc, cmax\n:\n0 . . N-1\nbegin\nrcv data(s, t, d) from qw →\nif MD(s; t; sq[0]) = d ∨MD(s; t; sq[1]) = d →\nif s < exp ∧c ̸= cmax →\n{reject message; report an adversary}\nc := (c + 1)modN\nexp ≤s ∨c = cmax →\n{accept message}\nexp := s + 1\nif c ̸= cmax →\nc := (c + 1)modN\n" }, { "page_number": 233, "text": "222\n■\nSecurity in Wireless Mesh Networks\nc = cmax →\nc := 0;\ncmax := RANDOM(0, N −1)\nfi\nfi\nMD(s; t; sq[0]) ̸= d ∧MD(s; t; sq[1]) ̸= d →\n{report an adversary}\nskip\nfi\ntrue →\n{p receives data(s, t, d) from mesh router other than q and}\n{checks that its message digest is correct and}\n{its sequence number is within range}\nRTMSG\ntrue →\n{either p receives data(t) from an adjacent host or}\n{p generates the text t for the next data message}\nRTMSG\nend\nThe first and second actions of process ps have a statement RTMSG that\nis defined as follows:\nif NXT(t) = p →\n{accept message}\nskip\nNXT(t) = q →\nd := MD(nxt; t; sp);\nsend data(t, d) to qs;\nnxt := nxt + 1\nNXT(t) ̸= p ∧NXT(t) ̸= q →\n{compute next soft sequence number s for sending data to NXT(t);}\n{compute d as message digest of concatenation of s, t}\n{and the secret for sending data to NXT(t);}\n{forward data(s, t, d) to router NXT(t)}\nskip\nfi\nTo verify the correctness of the strong integrity check protocol, use the\nstate transition diagram of this protocol in Figure 6.4, which considers only\nthe channel from process qs to process ps. (The channel from ps to qs\nand the channels from ps to any other strong integrity check process in an\nadjacent router of p can be verified in the same way.) This diagram has\n" }, { "page_number": 234, "text": "Hop Integrity in Wireless Mesh Networks\n■\n223\nsq in qs = sq[0] in ps\ns < exp in ps)\nc = cmax in ps)\nc ≠ cmax in ps)\n(Head(data(s, t, d))\nsq in qs = sq[1] in ps\nI =\nwhere\nU.0 =\nM.0 =\nP.0 =\nP.1 =\nI\nS:data\nS:data\nS:data\nS:data\nR:data &\ndiscard\nR:data &\naccept\nR:data & accept\nR:data &\ndiscard\nP:data\nP:data\nP:data\nM:data\nP.0\nU.0\nP.1\nM.0\nL:data\n(∀data(s, t, d) message in ch.qs.ps,\n(∀data(s, t, d) message in ch.qs.ps,\n(∀data(s, t, d) message in ch.qs.ps,\n(∀data(s, t, d) message in ch.qs.ps,\n(Head(data(s, t, d))\n( Head(data(s, t, d))\n( Head(data(s, t, d))\n(Head(data(s, t, d))\nd = MD(s; t; sq))\nd = MD(s; t; sq)\nd ≠ MD(s; t; sq)))\ns < exp in ps)\ns ≥ exp in ps))\nd = MD(s; t; sq)\nd = MD(s; t; sq)\n∧\n∧\n∧\n∧\n∧\n∧\n∧\nI ∧\nI ∧\nI ∧\n∨\n⇒\n⇒\n⇒\n⇒\n⇒\nFigure 6.4\nState transition diagram of the strong integrity check protocol.\nfour nodes that represent all possible reachable states of the protocol. Every\ntransition in the diagram stands for either a legitimate action (of process ps\nor process qs) or an illegitimate action of the adversary.\nNote that because the strong integrity check protocol operates below\nthe secret exchange protocol in the protocol stack, the assertion can be\nmade that (sq in qs = sq[0] in ps ∨sq in qs = sq[1] in ps) is an invariant in\nevery state of the strong integrity check protocol; this invariant is denoted\nas I in the specification in Figure 6.4.\nInitially, the protocol starts at state U.0. At state U.0, two legitimate ac-\ntions, namely, the send action in qs that sends a data message and the\nreceive action in ps that receives a data message, can be executed. Execut-\ning either one of the two actions at state U.0 keeps the protocol in state U.0.\nState U.0 is the only good state in the strong integrity protocol. The set\nof transitions that leads the protocol from state U.0 to state U.0 constitutes\nthe good cycle of the protocol. If only legitimate actions of processes ps\n" }, { "page_number": 235, "text": "224\n■\nSecurity in Wireless Mesh Networks\nand qs are executed, the protocol will stay in this good cycle indefinitely.\nNext, the bad effects caused by the actions of an adversary and how the\nprotocol can recover from these effects will be discussed.\nFirst, the adversary can execute a message loss action at state U.0. If\nthe adversary executes a message loss action at state U.0, the predicate\nthat for every data message data(s, t, d) in the channel from qs to ps,\nd = MD(s; t; sq), still holds. Therefore, the protocol stays at state U.0.\nSecond, the adversary can execute a message modification action at\nstate U.0 causing the protocol to move to state M.0. The receive and discard\naction executed by ps at state M.0 leads the protocol back to state U.0.\nThird, the adversary can execute a message replay action at state U.0.\nThere are two cases to consider. First, if the replayed message data(s, t, d) is\ntoo old such that the secret used to compute the message digest is different\nfrom the current value of constant sq in process qs, then the protocol moves\nto state M.0, and later returns to state U.0 as discussed above. Second, if\nthe replayed message data(s, t, d) is recent such that the secret used to\ncompute the message digest is equal to the current value of constant sq in\nprocess qw, then the protocol moves either to state P.0 or to state P.1. With\na high probability of (cmax−1)/cmax, the protocol moves to state P.0, and\nthe replayed message will be received and discarded by ps because the\nvalue of field s in the message indicates that the message is replayed. With\na probability of 1/cmax, the protocol moves to state P.1, and the replayed\nmessage will be received and accepted. In both cases the protocol returns\nto state U.0.\nFrom the state transition diagram, it is clear that each illegitimate action\nby the adversary will eventually lead the protocol back to U.0, which is a\ngood state. Once the protocol is in a good state, the protocol can progress\nin the good cycle. Moreover, if the adversary replays a recent data message,\nthe replayed message will be detected and discarded with high probability\n(cmax −1)/cmax. Hence the following two theorems about the strong\nintegrity check protocol are proved:\nTheorem 5\nIn the absence of an adversary, a network that executes the strong integrity\ncheck protocol follows the good cycle, consisting of a single transition from state\nU.0 to state U.0, and will stay in this good cycle indefinitely.\nTheorem 6\nIn the presence of an adversary, a network that executes the strong integrity\ncheck protocol will converge to the good cycle in a finite number of steps after\nthe adversary finishes executing any number of message loss or message mod-\nification actions. This network will also converge to the good cycle in a finite\nnumber of steps after the adversary finishes executing any number of message\nreplay actions.\n" }, { "page_number": 236, "text": "Hop Integrity in Wireless Mesh Networks\n■\n225\nThe protocols used by the weak hop integrity protocol and the strong\nhop integrity protocol have several novel features that make them cor-\nrect and efficient. First, whenever the secret exchange protocol attempts to\nchange a secret, it keeps both the old secret and the new secret until it is\ncertain that the integrity check of any future message will not be computed\nusing the old secret. Second, the integrity check protocol computes a digest\nat every router along the message route so that the location of any occur-\nrence of message modification can be determined. Third, the soft sequence\nnumber protocol makes the strong hop integrity protocol tolerate any loss\nof synchronization between any two adjacent routers.\n6.6\nConclusion and Open Issues\nThis chapter has presented scenarios of message insertion attacks and mes-\nsage replay attacks that may result in denial-of-service attack to wireless\nmesh networks, and introduces the hop integrity concept, which aims to\nprovide protection against these attacks. Then, the chapter presented the\nthree components of the hop integrity protocol suite for wireless mesh\nnetworks, namely, the initial authentication protocol, the secret exchange\nprotocol, and the integrity check protocol. Together, they provide hop in-\ntegrity to wireless mesh networks and their correctness is verified by state\ntransition diagrams.\nThere are a few open issues that are worth mentioning. The first open\nissue is on strategic deployment of hop integrity. Hop integrity protocols\nare open to incremental deployment, and the security they provide in-\ncreases with the number of pairs of hop integrity-equipped mesh routers\nbecause an adversary will have less venues to apply its attacks. However,\ndue to hardware/software compatibility and efficiency consideration, it may\nbe worthwhile to consider a strategic deployment scheme. For example, a\nfew hotspots in the network can be required to install static hop integrity,\nin which hop integrity is always turned on; other spots in the network\ncan install dynamic hop integrity, in which hop integrity is randomly on\nand off.\nThe second open issue is about interoperability between different wire-\nless mesh networks. The initial authentication protocol is designed for mesh\nrouters that belong to the same domain. For mesh routers from different\ndomains to execute these protocols, the certificates of the involved domains\nneed to be integrated.\nThe third open issue is about integrity in MAC and PHY layers. Wireless\nmesh networks are vulnerable to security attacks at various layers. Although\nthe protocols presented in this chapter address the integrity problem at\nnetwork layer, the same issue at the lower MAC and PHY layers is still an\nopen problem.\n" }, { "page_number": 237, "text": "226\n■\nSecurity in Wireless Mesh Networks\nReferences\n[1]\nC.-T. Huang and M. G. Gouda, Hop Integrity in the Internet, Springer,\nDecember 2005.\n[2]\nM. G. Gouda, Elements of Network Protocol Design, Wiley, April 1998.\n[3]\nY. Zhang, J. Luo, and H. Hu, Eds., Wireless Mesh Networking: Architectures,\nProtocols, and Standards, Auerbach Publications, Boca Raton, FL, 2006.\n[4]\nS. Kent and R. Atkinson, Security Architecture for the Internet Protocol, RFC\n2401, November 1998.\n[5]\nS. Kent and R. Atkinson, IP Authentication Header, RFC 2402, November\n1998.\n[6]\nS. Kent and R. Atkinson, IP Encapsulating Security Payload (ESP), RFC 2406,\nNovember 1998.\n[7]\nM. G. Gouda, Elements of Network Protocol Design, Wiley, April 1998.\n[8]\nM. G. Gouda, E. N. Elnozahy, C.-T. Huang, and T. M. McGuire, Hop integrity\nin computer networks, IEEE/ACM Transactions on Networking, Vol. 10, No.\n3, June 2002.\n[9]\nC.-T. Huang, Hop Integrity: A Defense against Denial-of-Service Attacks,\nPh.D. dissertation, Department of Computer Sciences, The University of\nTexas at Austin, August 2003.\n[10]\nC.-T. Huang and M. G. Gouda, Hop Integrity in the Internet, Springer,\nDecember 2005.\n[11]\nT. Dierks and C. Allen, The TLS Protocol Version 1.0, RFC 2246, January\n1999.\n[12]\nKerberos: The Network Authentication Protocol, http://web.mit.edu/\nKerberos/.\n[13]\nC. Rigney, S. Willens, A. Rubens, and W. Simpson, Remote Authentication\nDial In User Service (RADIUS), RFC 2865, June 2000.\n[14]\nIETF Public-Key Infrastructure (X.509) (pkix) Charter, http://www.ietf.\norg/html.charters/pkix-charter.html.\n[15]\nR. Housley, W. Polk, W. Ford, and D. Solo, Internet X.509 Public Key\nInfrastructure Certificate and Certificate Revocation List (CRL) Profile, RFC\n3280, April 2002.\n[16]\nR. L. Rivest, The MD5 Message-Digest Algorithm, RFC 1321, 1992.\n[17]\nNIST, FIPS PUB 180-1: Secure Hash Standard, April 1995.\n[18]\nH. Krawczyk, M. Bellare, and R. Canetti, HMAC: Keyed-Hashing for\nMessage Authentication, RFC 2104, February 1997.\n" }, { "page_number": 238, "text": "Chapter 7\nPrivacy Preservation\nin Wireless Mesh\nNetworks1,2\nTaojun Wu, Yuan Xue, and Yi Cui\nContents\n7.1\nIntroduction ......................................................... 228\n7.2\nPrivacy Preserving Architecture ..................................... 230\n7.3\nPrivacy Modeling in WMNs ......................................... 232\n7.3.1\nNetwork Model .............................................. 232\n7.3.2\nTraffic Entropy ............................................... 233\n7.3.2.1\nBasic Definition .................................... 233\n7.3.2.2\nMutual Information ................................. 235\n7.4\nPenalty-Based Routing Algorithm................................... 236\n7.5\nExperimental Results ................................................ 239\n7.5.1\nSimulation Setup ............................................. 239\n7.5.2\nTraffic Entropy and Mutual Information .................... 240\n1 This work was supported in part by TRUST (The Team for Research in Ubiquitous\nSecure Technology), which receives support from the National Science Foundation\n(NSF award number CCF-0424422) and the following organizations: Cisco, ESCHER,\nHP, IBM, Intel, Microsoft, ORNL, Pirelli, Qualcomm, Sun, Symantec, Telecom Italia,\nand United Technologies.\n2 Copyright c⃝IEEE, 2006. This is an extension of the short paper published in IEEE\nInternational Symposium on a World of Wireless, Mobile and Multimedia Networks\n(WoWMoM), 2006.\n227\n" }, { "page_number": 239, "text": "228\n■\nSecurity in Wireless Mesh Networks\n7.5.3\nWhich Nodes Have More Mutual Information? ............. 240\n7.5.4\nTrade-Off between Performance Degradation\nand Traffic Privacy ........................................... 244\n7.6\nCollusion Analysis ................................................... 247\n7.6.1\nProblem Description ......................................... 248\n7.6.2\nColluded Traffic Mutual Information ........................ 249\n7.6.3\nSimulation Results ........................................... 250\n7.6.3.1\nTraffic Curves ....................................... 251\n7.6.3.2\nColluded Traffic Mutual Information:\nSingle Pair of Observers............................ 252\n7.6.3.3\nColluded Traffic Mutual Information:\nMultiple Pairs of Observers ........................ 253\n7.7\nRelated Work ........................................................ 255\n7.8\nConclusion........................................................... 257\nReferences................................................................. 258\nMulti-hop wireless mesh networking (WMN) has attracted increasing\nattention and deployment as a low-cost approach to provide last-mile broad-\nband Internet access. Privacy is a critical issue in WMN, as traffic of an end\nuser is relayed via multiple wireless mesh routers. Due to the unique char-\nacteristics of WMN, the existing solutions for the Internet are either ineffec-\ntive at preserving privacy of WMN users, or will cause severe performance\ndegradation.\nIn this chapter, we propose a lightweight privacy preserving solution\naimed to achieve well-maintained balance between network performance\nand traffic privacy preservation. At the center of this solution is an inform-\nation-theoretic metric called “traffic entropy,” which quantifies the amount\nof information required to describe the traffic pattern and to characterize the\nperformance of traffic privacy preservation. We further present a penalty-\nbased shortest path routing algorithm that maximally preserves traffic\nprivacy by minimizing the mutual information of “traffic entropy” observed\nat each individual relaying node, meanwhile controlling performance degra-\ndation within the acceptable region. Extensive simulation study proves the\nsoundness of our solution and its resilience to cases when two malicious\nobservers collude.\n7.1\nIntroduction\nRecently, multi-hop WMN has attracted increasing attention and deploy-\nment as a low-cost approach to provide last-mile broadband Internet\naccess [2–5]. In WMN, each client accesses a stationary wireless mesh router.\nMultiple mesh routers communicate with one another to form a multi-hop\nwireless backbone that forwards user traffic to a few gateways connected to\n" }, { "page_number": 240, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n229\nthe Internet. Some perceived benefits of WMN include enhanced resilience\nagainst node failures and channel errors, high data rates, and low costs\nin deployment and maintenance. For such reasons, commercial WMNs are\nalready deployed in some U.S. cities (like Medford and Chaska). Even large\ncities are planning to deploy citywide WMNs as well [1].\nHowever, to further widen the deployment of WMN and enable it as a\ncompetitive player in the market of broadband Internet access, the issue of\nprivacy must be addressed. Privacy has been a major concern of Internet\nusers [12]. It is a particularly critical issue in the context of WMN-based\nInternet access, where users’ traffic is forwarded via multiple mesh routers.\nIn a community mesh network, this means that the traffic of a residence\ncan be observed by the mesh routers residing at its neighbors. Despite the\nnecessity, limited research has been conducted toward privacy preservation\nin WMN.\nThis motivates us to investigate the privacy preserving mechanism in\nWMN. There are mainly two privacy issues: data confidentiality and traffic\nconfidentiality.\n■\nData confidentiality: It is obvious that data content reveals user pri-\nvacy on what is communicated. Data confidentiality aims to protect\nthe data content and prevent eavesdropping by intermediate mesh\nrouters. Message encryption is a conventional approach for data\nconfidentiality.\n■\nTraffic confidentiality: Traffic information such as who the users are\ncommunicating with, when and how frequently they communicate,\nand the amount and the pattern of traffic, also reveals critical privacy\ninformation. The broadcasting nature of wireless communication\nmakes acquiring such information easy. In a WMN, attackers can\nconduct traffic analysis at mesh routers by simply listening to the\nchannels to identify the “ups and downs” of the target’s traffic. While\ndata confidentiality can be achieved via message encryption, it is\nmuch harder to preserve traffic confidentiality. In this chapter we\nfocus on the user traffic confidentiality issue and study the problem\nof traffic pattern concealment.\nWe aim at designing a lightweight privacy preserving mechanism for\nWMN which is able to balance the traffic analysis resistance and the band-\nwidth cost. Our mechanism makes use of the intrinsic redundancy of WMN,\nwhich is able to provide multiple paths for data delivery. By intuition, if the\ntraffic from the source (i.e., gateway) to the destination (i.e., mesh router)\nis split to many paths, then all the relaying nodes3 along the paths could\n3 In this chapter we use the following terms interchangeably: wireless mesh router,\nintermediate relaying node, and wireless node.\n" }, { "page_number": 241, "text": "230\n■\nSecurity in Wireless Mesh Networks\nonly observe a portion of the entire traffic. Moreover, if the traffic is split\nin a random way both spatially and temporally, then an intermediate node\nhas limited knowledge to figure out the overall traffic pattern. Thus the\ntraffic pattern is concealed.\nBased on this intuition, we seek a routing scheme which routes data\nsuch that the statistical distributions of the traffic observed at intermediate\nrelaying nodes are independent from the actual traffic from the source\nto the destination. To achieve this goal, we first define an information-\ntheoretic metric, traffic entropy, which quantifies the amount of information\nrequired to describe the traffic pattern. Then we present a penalty-based\nrouting algorithm, which aims to minimize the mutual information of traffic\nentropy observed at each relaying node, meanwhile controling the network\nperformance degradation under the acceptable level.\nConsidering the possibility of collusion, we evaluate our scheme under\na situation when two observers exchange their knowledge about the same\ndestination. We measure this shared knowledge as “colluded traffic mutual\ninformation” and our simulation results show that our scheme is still viable\nin case of two colluding eavesdroppers.\nThe rest of this chapter is organized as follows. In Section 7.2, we present\nthe overall architecture for privacy preservation in WMN. Section 7.3 and\nSection 7.4 focus on the traffic privacy issue. In particular, Section 7.3\npresents the model to quantify the performance of traffic privacy pre-\nservation, and Section 7.4 presents the routing algorithm. The proposed\nprivacy preserving solution is evaluated via extensive simulation study in\nSection 7.5. Section 7.6 discusses the collusion problem possible with mali-\ncious traffic observers and its impact on our proposed scheme. Section 7.7\nsummarizes background knowledge and related work. Section 7.8 con-\ncludes the chapter and points out the future directions.\n7.2\nPrivacy Preserving Architecture\nWe consider a multi-hop WMN shown in Figure 7.1. In this network, client\ndevices access a stationary wireless mesh router at its residence. Multiple\nmesh routers communicate with one another to form a multi-hop wireless\nbackbone that forwards user traffic to the gateway which is connected to\nthe Internet.\nTwo privacy aspects are considered in this architecture. Data confiden-\ntiality aims to protect the data content from eavesdropping by the interme-\ndiate mesh routers. Traffic confidentiality prevents the traffic analysis attack\nfrom the mesh routers, which aims at deducing the traffic information such\nas who the user is communicating with and the amount and the pattern of\ntraffic. Our privacy preserving architecture aims to protect the privacy of\n" }, { "page_number": 242, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n231\nInternet\nGateway g\nKUg, KRg\nKUi, for all mesh router i \nClient device\nMesh router i\nKUi, KRi, KUg\n(g, a, b, c, e, i) s, d \nClient \nSource route\nEncrypted packet\nHigher layer data\na\nb\nc\ne\ns\nFigure 7.1\nPrivacy preserving architecture for wireless mesh network.\neach wireless mesh router, the basic routing unit in WMN. The architecture\nconsists of the following functional components:\n■\nKey distribution: In this architecture, each mesh node, as well as\nthe gateway, has a pair of public and private keys (KU, KR). The\ngateway maintains a directory of certified public keys of all mesh\nnodes, and each mesh node has a copy of the public key of the\ngateway KU g. The public key KU i of mesh node i and KU g are\nused to establish the shared secret session key KSgi, which is used\nto encrypt the messages between them.\n■\nMessage encryption: Let M be the IP packet sent from a source s\nin the Internet to a client d in the mesh network, and let i be the\nmesh router of client d. The whole IP packet M, which contains\nthe original source and destination address s and d, is encrypted at\ngateway g via the shared secret key KSgi: Me = E (KSgi, M). To route\nthe encrypted packet Me to its destination, the gateway prefixes the\nsource route from the gateway g to the router i to the packet. The\nencapsulated packet is then forwarded by relaying routers in WMN.\nLikewise, packets travelling in the reverse direction are treated the\n" }, { "page_number": 243, "text": "232\n■\nSecurity in Wireless Mesh Networks\nsame way. As the source address s and other higher-layer header\ninformation, such as port, are all encrypted, the relaying routers\nare unable to obtain the information on who the client of router\ni is communicating with and what type of application is involved.\nBecause encryption and decryption take place only at the gateway\nand the destination mesh router, much less computation is required,\nwhich is a desired feature in WMN.\n■\nRouting control: With source route in cleartext in an encapsulated\npacket, the intermediate mesh routers can still observe the amount\nand the pattern of the traffic of a particular mesh node i. To ad-\ndress this problem, our privacy preserving mechanism explores the\npath diversity of WMN and forwards packets between the gateway\nand the mesh node via different routes. Thus any relaying router\ncan only observe a portion of the whole traffic of this connection.\nIn Section 7.4, we detail the design of a penalty-based routing al-\ngorithm, which randomly selects a route for each individual packet\nsuch that the observed traffic pattern at each relaying node is inde-\npendent of the overall traffic. In our design, the gateway maintains\na complete topology of the WMN and computes the source routes\nbetween the destination mesh nodes and itself.\n7.3\nPrivacy Modeling in WMNs\n7.3.1\nNetwork Model\nWe model the WMN shown in Figure 7.1 as a graph G = {V, E}, where V is\nthe set of wireless nodes in WMN, and E is the set of wireless edges (x, y)\nbetween any two nodes x, y. Each node x maintains a logical connection\nwith the gateway node g. Node x receives data from the Internet via g. The\nsource and destination information of a packet is open to the relaying node.\nThe traffic pattern of x can be categorized into two types: incoming traffic\npattern and outgoing traffic pattern. In this paper, we mainly consider the\nfirst type.\nIf the traffic between s and x goes through only one route, then any\nrelaying node on this route can easily observe the entire traffic between\ng and x, thus violating its traffic pattern privacy. To avoid this problem, x\nmust establish multiple paths with g and distribute its traffic along these\npaths, such that any node can only get a partial picture of x ’s traffic pattern.\nHowever, the complete traffic pattern information of x could still be\nobtained by a single node in case of multi-path routing. In the example\nshown by Figure 7.2, g allocates the traffic to x via three disjoint routes\nby fixed proportion. Then for any node along any path, although only\nseeing one third of the flow, the observed traffic shape is isomorphic to the\noriginal one. Therefore, the traffic to x must be distributed along multiple\n" }, { "page_number": 244, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n233\nTime\nTraffic volume\nTotal traffic of x\nTraffic routed through a path\nFigure 7.2\nAn example of isomorphic traffic.\nroutes in a time-variant fashion, such that the traffic pattern observed at any\nnode is statistically deviant from the original pattern. The notations used in\nSection 7.3 are listed in Table 7.1.\n7.3.2\nTraffic Entropy\nWe propose to use information entropy as the metric to quantify the per-\nformance of a solution at preserving the traffic pattern confidentiality. In\nwhat follows, we consider two nodes x and y; x is the destination node of\nthe traffic from the gateway g to x, y is the observing node, which relays\npackets for x and also tries to analyze the traffic of x.\n7.3.2.1\nBasic Definition\nIdeally, we view the traffic of x as a continuous function of time, as shown\nin Figure 7.3. In practice, the traffic analysis is conducted by dividing time\nTable 7.1\nNotations Used in Section 7.3\nV\nWireless node set\nE\nEdge set\ng\nGateway node\nx\nDestination node\ny\nObserving node\nX\nRandom variable describing x’s traffic pattern\nY X\nRandom variable describing x’s traffic pattern observed by y\nH(X)\nEntropy of X\nH(Y X)\nEntropy of Y X\nI (Y X, X)\nMutual information between X and Y X\n" }, { "page_number": 245, "text": "234\n■\nSecurity in Wireless Mesh Networks\nTime\nTraffic volume\n……\nTotal traffic of x\nFigure 7.3\nSampling-based traffic analysis.\ninto equal-sized sampling periods, then measuring the amount of traffic in\neach period, usually in terms of number of packets, assuming the packet\nsizes are all equal. Therefore, as the first step, we discretize the continuous\ntraffic curve into piecewise approximation of discrete values, each denoting\nthe number of packets destined to x in a sampling period.\nNow, we use X as the random variable of this discrete value. Y X is the\nrandom variable representing the number of packets destined to x observed\nat node y in a sampling period. We denote P(X = i) as the probability\nthat the random variable X is equal to i (i ∈N ), i.e., the probability that\nnode x receives i packets in a sampling period. Likewise, P(Y X = j) is the\nprobability that Y X is equal to j ( j ∈R), i.e., j packets destined to x go\nthrough node y in a sampling period.\nThen the discrete Shannon entropy of the discrete random variable X is\nH(X ) = −\n\u0002\ni\nP(X = i) log2 P(X = i)\n(7.1)\nH(X) is a measurement of the uncertainty about the outcome of X.\nIn other words, it measures the information of node x ’s traffic, i.e., the\nnumber of bits required to code the values of X. H(X ) takes its maximum\nvalue when the value of X is uniformly distributed. On the other hand, if\nthe traffic pattern is CBR, then H(X ) = 0 because the number of packets\nat any sampling period is fixed.4\n4 This offers the information-theoretic interpretation for traffic padding: by flattening\nthe traffic curve with blank packets, the entropy of observable traffic is reduced to 0,\nwhich perfectly hides the information of the original traffic pattern.\n" }, { "page_number": 246, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n235\nSimilarly, we have the entropy for Y X as follows:\nH(Y X) = −\n\u0002\nj\nP(Y X = j) log2 P(Y X = j)\n(7.2)\n7.3.2.2\nMutual Information\nWe then define the conditional entropy of random variable Y X with respect\nto X as\nH(X|Y X) = −\n\u0002\nj\nP(Y X = j)\n\u0002\ni\npij log2 pij\n(7.3)\nwhere pij = P(X = i|Y X = j) is the probability that X = i given the\ncondition that Y X = j. H(X|Y X) can be thought of as the uncertainty\nremaining about X after Y X is known. The joint entropy of X and Y X can\nbe shown as\nH(X, Y X) = H(Y X) + H(X|Y X)\n(7.4)\nFinally, we define the mutual information between X and Y X as\nI (Y X, X) = H(X) + H(Y X) −H(X, Y X)\n= H(X) −H(X|Y X)\n(7.5)\nwhich represents the information we gain about X from Y X.\nBack to the example in Figure 7.2, let us assume that the observing\nnode y is located on one route destined to x. Because the traffic shape\nobserved at y is the same as x, at any sampling period, if Y X = j, then X\nmust equal to a fixed value i, making P(X = i|Y X = j) = 1. According to\nEquation (7.3), this makes the conditional entropy H(X|Y X) = 0. According\nto Equation (7.5), we have I (Y X, X) = H(X), implying that from Y X, we\ngain the complete information about X.\nOn the contrary, if Y X is independent from X, then the conditional\nprobability P(X = i|Y X = j) = P(X = i), which maximizes the condi-\ntional entropy H(X|Y X) to H(X). According to Equation (7.5), we have\nI (Y X, X) = 0,5 i.e., we gain no information about X from Y X.\nIn reality, because Y X records the number of a subset of packets des-\ntined to node x, it cannot be totally independent from the random vari-\nable X. Therefore, the mutual information should be valued between the\n5 By the definition of mutual information, I (Y X, X) ≥0, with equality if and only if X\nand Y are independent.\n" }, { "page_number": 247, "text": "236\n■\nSecurity in Wireless Mesh Networks\ntwo extremes discussed above, i.e., 0 < I (Y X, X) < H(X). This means that\nnode y can still obtain partial information of X ’s traffic pattern. However,\na good routing solution should minimize such mutual information as much\nas possible for any potential observing node. More formally, we should\nminimize\nmax\nY∈V−X I (Y X, X)\n(7.6)\nthe maximum mutual information that any node can obtain about X.\n7.4\nPenalty-Based Routing Algorithm\nIn this section, we propose a penalty-based routing algorithm to achieve our\ngoal of hiding the traffic pattern by exploiting the richness of available paths\nbetween two nodes in WMN. Specifically, we choose to adopt the source\nrouting scheme. Such a choice is enabled by the fact that one node can\neasily acquire the topology of the WMN it belongs to, which is mid-sized\n(within 100 nodes) and static.\nWhen designing the algorithm, we also keep in mind the need to\ncompromise between sufficient security assurance and acceptable system\noverhead. We would show in our algorithm that system performance is\nsatisfactory and security assurance is adequate.\nShown in Table 7.2, the algorithm operates in three phases: path pool\ngeneration, candidate path selection, and individual packet routing. The\nnotations used in this section are listed in Table 7.3.\nFirst, in the path pool generation phase, we try to generate a large set\nof diversified routing paths connecting the gateway g and the destination\nnode x, denoted as Spaths. The path generation algorithm is an iterated\nprocess of applying a modified version of Dijkstra’s algorithm. Here, each\nnode is assigned a penalty weight, and the weight of an edge is defined as\nthe weighted average of penalty weights of its two end nodes. The weight\n(or cost) of a path is defined as the sum of penalty weights of all edges\nconsisting this path. The algorithm runs in iterations. Initially, we set the\npenalty weight of each node as 1, then run Dijkstra’s algorithm to find the\nfirst shortest path from the gateway g to x. Next, we increase the penalty\nweight for each node on this found path. This will make these appeared\nnodes less competitive to other nodes in becoming components of next\npath. After this, the algorithm proceeds to the next iteration, generating\nthe second path, and all nodes appearing on the second path are penal-\nized through increasing their weights. This process goes on until enough\nnumbers of paths are found.\nSecond, in the candidate path selection phase, we try to choose a\ncombination of diversified routing paths, a subset of paths from the set\n" }, { "page_number": 248, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n237\nTable 7.2\nPenalty-Based Routing Algorithm\n/*Penalty-Based Shortest Path*/\nP BSP(Snode, Dnode)\nFor each node v ∈V\nd[v] ←∞\nFor each node v ∈V\npr ev[v] ←∞\nFor each node v ∈V\nvisited[v] ←0\nd[SNode] ←0\nRepeat\nGet unvisited vertex v with the least d[v]\nIf d[v] ≥∞, Then v unreachable\nElse visited[v] ←1\nFor all v’s neighbors w\nE dgePenalty = α[pow(γ, (w.tag))] + β(v.tag)\nIf d[w] > d[v] + E dgePenalty\nd[w] ←d[v] + E dgePenalty\npr ev[w] ←v\nUntil visited[v] = 1, ∀v ∈V\n/*Generate Spaths For Each g −x Pair*/\nGenPath()\nFor All Non-Gateway Nodes x\nFor each node v ∈V\nv.tag ←1\nRepeat\nPBSP(g, x)\nGet new g −x path Pnew from vector pr ev[]\nStore Pnew in Spaths\nFor all nodes v on Pnew\nv.tag ←v.tag + 1\nUntil PathPoolSize paths found.\n/*Select Sselected For Each g −x Pair*/\nSel Path()\nRepeat\nr nd = r and() mod PathPoolSize\nselect r ndth path from Spaths\nUntil Sel PathNum paths selected\n/*Decide path for arriving packet*/\nRoutePkt(Snode, Dnode)\nPackets[Dnode] ←Packets[Dnode] + 1\nr ndpath = r and() mod Sel PathNum\nroute packet along the r ndpathth path from Sselected\nIf Packets[Dnode] > ReSel PathCnt\nPackets[Dnode] ←0\nSelPath()\n" }, { "page_number": 249, "text": "238\n■\nSecurity in Wireless Mesh Networks\nTable 7.3\nNotations Used in Section 7.4\nv, w\nNode\nv.tag\nNumber of times v is included by a path\nα\nFactor to slow down penalty rate\nβ\nFactor to avoid many identical paths in beginning\nstages of path generation\nγ\nBase of exponential penalty function\nd[]\nPenalty vector for every node\npr ev[]\nVector to store Pnew reversely\nPackets[]\nVector to store number of arrived packets for every node\nSpaths, denoted as Sselected. The paths in Sselected are selected randomly from\nSpaths. After each choice of a path into Sselected, the probability factor of\nthat path is decreased to lower the chance of multiple identical paths exist-\ning in Sselected. Sselected is changed and renewed corresponding to network\nactivities.\nThird, in the packet routing phase, we choose randomly from Sselected\none path for each packet and increase the counter for the selected path\nsubset Sselected. This Sselected path subset expires after counter reaches its\npredetermined threshold. Then Sselected is renewed by calling the second\nphase again.\nBecause packets are assigned a randomly chosen path, and all these\ncandidate paths are designed to be disjoint, the chance that packets are\nrouted in similar paths is small. Our experiment results further confirm this\nintuition.\nThis algorithm is designed to balance the needs of routing performance\n(finding paths with smallest hop count) and preserving traffic pattern pri-\nvacy (finding disjoint paths). The penalty weight update function serves\nas the tuning knob to maneuver the algorithm between these two con-\ntradictory goals. During the initialization, when the penalties of all nodes\nare equal, the path found by the algorithm is indeed shortest in terms of\nhop count. As a node is chosen by more routes, its penalty weight mono-\ntonically increases, making it less likely to be chosen again. Thus, as the\nalgorithm proceeds, the newly chosen paths (shortest in terms of its aggre-\ngate penalty weight) become more disjoint from existing paths, but longer\nin terms of hop count. The pace of such shift from “smallest hop-count\npath” to “disjoint path” is controlled by how fast the penalty weight update\nfunction grows. Our experiment results confirm this reasoning. Finally, by\nrandomly assigning packets along different paths, the algorithm maximally\ndisturbs the traffic pattern of any g −x pair.\nAlthough penalty-based routing has been used in existing literature [8],\nwe are using it for different objects. Their links were penalized for losses\n" }, { "page_number": 250, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n239\n0\n100\n200\n300\n400\n500\n600\n0\n100\n200\n300\n400\n500\n600\nY position\nX position\n 0\n 1\n 2\n 3\n 4\n 5\n 6\n 7\n 8\n 9\n10\n11\n12\n13\n14\n15\n16\n17\n18\n19\n20\n21\n22\n23\n24\n25\n26\n27\n28\n29\nGateway\nFigure 7.4\nExperimental topology.\nor malicious behavior while our approach applies it to avoid using links\nrepeatedly to get better path diversity.\n7.5\nExperimental Results\n7.5.1\nSimulation Setup\nWe base our simulation on a randomly generated topology (Figure 7.4)\n(600 × 600) with 30 nodes. The effective distance between two nodes is\nset to be 250. The whole process of simulation consists of 400,000 logical\nticks. In each single tick, a packet is generated at gateway node 0 and its\ndestination is randomly decided to be one of the other 29 nodes. To better\nsimulate real network traffic, we set the probability of 0.05 that, at one tick,\nno packet is generated, i.e., idle probability. The distance delay factor is\nchosen to be 0.003 tick and the hop delay factor is decided as 0.05 tick. We\napproximate hop delay at any node by multiplying the hop delay factor\nwith its usage count by all paths chosen initially.\nWith a relatively small node set, we choose 50 as our PathPoolSize\nand 5 as Sel PathNum. The selected path subset Sselected for any destination\nnode is renewed after sending 50 packets to that node. To obtain multi-\nple diversified paths with Dijkstra’s algorithm more quickly, we introduce\nthe exponential penalty function on the tag of one node and use γ as\n" }, { "page_number": 251, "text": "240\n■\nSecurity in Wireless Mesh Networks\nthe base of exponential function when deciding on which edge to include\nthe candidate path. To slow down the growing rate of exponential penalty\nfunction, we multiply the exponential function with a factor α when cal-\nculating E dgePenalty. To avoid getting too many identical paths in the\nbeginning stages, we amplify the influence of another node by multiplying\nthe tag of another node with β. The penalty parameters α, β, γ are chosen\nto be 0.5, 15, and 1.85, respectively.\n7.5.2\nTraffic Entropy and Mutual Information\nThe total 400,000 ticks is divided into 20 periods. Each period is then di-\nvided into 50 intervals and one interval is 400 ticks long. Within each in-\nterval, for each destination node x, we count the number of packets that\nall other nodes y have relayed for x. Then for each period, we indepen-\ndently calculate the traffic entropies H(X), H(Y X) and mutual information\nI (Y X, X) based on their definitions in Section 7.3.2.\nDue to the space limit, we only show part of our results. Among all\nnodes in the network, we choose two sets of nodes. Nodes in the first\nset {1, 6, 11, 15, 23, 24, 25, 29} are close to (two to three hops) the gate-\nway node 0. Nodes in the second set {2, 3, 7, 16, 17, 28} are at the edge\nof the network, four to five hops away from the gateway. We choose two\nrepresentative nodes, 1 and 16, out of each set.\nFigure 7.5 shows the variance of traffic entropy and mutual information\nalong the time. In Figure 7.5 (a), H(1−1) denotes the traffic entropy of node\n1. H(23−1) denotes the traffic entropy of node 23 based on its observation\non node 1. M I (23−1, 1−1) denotes the mutual information node 23 shares\nwith node 1. The same notation rules apply for Figure 7.5 (b), where node\n16 is the destination and 9 is the observer. In both pictures, the observing\nnode only shares 40 percent or less of information about the observed\ndestination node at any sampling period.\nThis observation is further confirmed in Figure 7.6, where we plot the\ntime-variant mutual information that destinations 1 and 16 share with other\nrandomly chosen observing nodes. These results show that with our algo-\nrithm, the destination node is able to consistently limit the proportion of\nmutual information it shares with the observing nodes.\n7.5.3\nWhich Nodes Have More Mutual Information?\nIn Figure 7.7(a), we calculate the time-averaged mutual information for all\nobserving nodes with respect to the destination node 1, and sort them in\nthe ascending order. Here, we observe an almost linearly-growing curve\nexcept at its head and tail. For nodes at the head of the curve, their mutual\ninformation is 0 because they lie at the outer rim of the network, hence\nare not chosen by our routing algorithm to relay traffic for node 1. At the\n" }, { "page_number": 252, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n241\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic information entropy\nNumber of periods\nH (23–1)\nH (1–1)\nMI (23–1, 1–1)\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic information entropy\nNumber of periods\nH(9–16)\nH(16–16)\nMI(9–16, 16–16)\n(a) Destination: node 1, observer: node 23\n(b) Destination: node 16, observer: node 9\nFigure 7.5\nTraffic entropy along time (single observer, γ = 1.85).\ntail of the curve is destination node 1, whose mutual information is actu-\nally the traffic entropy of its own. In Figure 7.7 (b), we observe the same\nphenomenon for destination 16, except at the head of the curve. This is\nbecause its network location is at the opposite end of the gateway, making\nevery node of the network to be its candidate relaying node.\nThis leads us to investigate if such distribution of mutual information\nis related with any other factors. We tried to connect mutual information\nof each node with certain metrics, such as its distance to the destination,\nbut failed to find any causal relationship. We then sort observing nodes\n" }, { "page_number": 253, "text": "242\n■\nSecurity in Wireless Mesh Networks\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic information entropy\nNumber of periods\nH(1–1)\nMI(29–1, 1–1)\nMI(4–1, 1–1)\nMI(26–1, 1–1)\nMI(7–1, 1–1)\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic information entropy\nNumber of periods\nH(16–16)\nMI(26–16, 16–16)\nMI(2–16, 16–16)\nMI(1–16, 16–16)\nMI(10–16, 16–16)\n(a) Destination: node 1, observers: node 4, 7, 26, 29\n(b) Destination: node 16, observers: node 1, 2, 10, 26\nFigure 7.6\nTraffic\nentropy in different sampling periods (multiple observers,\nγ = 1.85).\nbased on the averaged relayed traffic (average number of packets each\nnode relays in a sampling period) on a log-log scale, and find the linear\ndistribution as shown in Figure 7.8.\nObviously, such a power-law correlation tells us that the more traffic an\nobserving node relays for a destination node, the more mutual information\ncan be obtained about its traffic entropy. Furthermore, it gives us one way\nto experimentally quantify the relationship of these two metrics. Let T be\nthe amount of traffic relayed and I be the mutual information; then their\n" }, { "page_number": 254, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n243\n0 \n0.5 \n1 \n1.5 \n2 \n2.5 \n3 \n3.5 \n4 \n0 \n5 \n10 \n15 \n20 \n25 \n30 \nTraffic information entropy \nNode \n(a)\n3 7 16 17 27 28 \n10 2 9 22 26 8 29 25 5 15 11 24 4 13 20 21 6 14 12 23 18 19 \n 1 \nAverage mutual information \n 0\n 0.5\n 1\n 1.5\n 2\n 2.5\n 3\n 3.5\n 4\n 0\n 5\n 10\n 15\n 20\n 25\n 30\nTraffic Information Entropy\nNode\n151310\n 7 11 1 29202224 8 212527 5 17 3 281814191223 9 2 26 4 6\n16\nAverage Mutual Information\n(b)\nFigure 7.7\nSorted traffic mutual information. (a) Destination: node 1 (γ = 1.85);\n(b) Destination: node 16 (γ = 1.85).\npower-law relationship can be written as\nI = aT k\n(7.7)\nwhere a is the constant of proportionality and k is the exponent of the\npower law, both of which can be measured from Figure 7.8. If k < 1, then\nthe mutual information of an observing node grows in a sub-linear fashion\nas the amount of its relayed traffic increases, and in a super-linear fashion\n" }, { "page_number": 255, "text": "244\n■\nSecurity in Wireless Mesh Networks\n 1\n 0.1\n 1\n 10\nTraffic Information Entropy\nAverage Relayed Traffic\n 0\n 2\n 4\n 6\n 8\n10\n12\n1418\n20\n22\n24\n26\nAverage Mutual Information\n(a)\n 1\n 0.1\n 1\n 10\nTraffic Information Entropy\nAverage Relayed Traffic\n 0\n 2 4\n 6\n 8\n10\n1214\n16\n18\n20\n22 24\n26\n28\nAverage Mutual Information\n(b)\nFigure 7.8\nPower law correlation of mutual information and amount of traffic re-\nlayed. (a) Destination: node 1 (γ = 1.85); (b) Destination: node 16 (γ = 1.85).\notherwise. From what we have in Figure 7.8 and the same results for other\ndestination nodes, k < 1. This means that each time to make its mutual\ninformation further grows with the same increment, an observing node has\nto relay more and more traffic.\n7.5.4\nTrade-Off between Performance Degradation\nand Traffic Privacy\nFinally, we study the performance trade-off of our algorithm by tuning its\nexponential penalty function base γ . The performance degradation intro-\nduced by our algorithm is captured by the average hop ratio. For each\n" }, { "page_number": 256, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n245\n 0.95\n 1\n 1.05\n 1.1\n 1.15\n 1.2\n 1.25\n 1.3\n 1.35\n 1.4\n 1.45\n 1.5\n1.12\n1.31\n1.57\n1.85\n2.06\n2.59\nRatios\nParameter γ\nNode 1\nNode 6\nNode 11\nNode 15\nNode 17\nNode 24\nNode 29\n(a)\n 0.95\n 1\n 1.05\n 1.1\n 1.15\n 1.2\n 1.25\n 1.3\n 1.35\n 1.4\n 1.45\n 1.5\n1.12\n1.31\n1.57\n1.85\n2.06\n2.59\nRatios\nParameter γ\nNode 2\nNode 3\nNode 7\nNode 16\nNode 23\nNode 25\nNode 28\n(b)\nFigure 7.9\nAverage hop ratio. (a) Hop ratio of nodes of first set; (b) Hop ratio of\nnodes in the second set.\ngateway-destination pair g −x, this metric is defined as the ratio between\nthe average number of hops a packet goes through using our algorithm and\nthe number of hops of the shortest path between g and s. From Figure 7.9,\nwe can see that the average hop ratio increases as γ increases. The direct\nneighbors of the gateway are less sensitive to the change of γ , like node 6\nin Figure 7.9(a) and node 23 in Figure 7.9(b).\nIn Figure 7.10 and Figure 7.11 we find that under shortest path routing,\nthe mutual information of a node is 0 if it is not on the path to destina-\ntion node. Otherwise, the mutual information node is much higher than\nthe case of our algorithm. Also worth noting is that increasing of γ has\na different impact on different nodes, depending on distance to gateway,\ndestination, and location in the WMN. Take nodes 12 (Figure 7.10) and 6\n(Figure 7.11) for example, because they lie near to the gateway node and\n" }, { "page_number": 257, "text": "246\n■\nSecurity in Wireless Mesh Networks\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\n0\n0.2\n0.4\n0.6\n0.8\n1\n1.2\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\n(b) Observer: node 15\n(a) Observer: node 12\n0\n0.1\n0.2\n0.3\n0.4\n0.5\n0.6\n0.7\n0.8\n0.9\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\n0\n0.2\n0.4\n0.6\n0.8\n1\n1.2\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\n(d) Observer: node 26\n(c) Observer: node 22\nFigure 7.10\nTraffic mutual information under different penalty parameters (desti-\nnation: node 1).\nare relatively centrally situated, their observed mutual information varies\nlittle with respect to the change of γ . Whereas for node 22 (Figure 7.10),\nwhich is far away from destination node 1 and on the edge of the WMN,\nmutual information shared with node 1 increases with the growth of γ ,\nindicating more traffic is routed through farther nodes. This tendency of\nrouting packets from farther nodes leads to a higher average number of\nhops, which is confirmed by our analysis about average hop ratio. How-\never, traffic mutual information tends to decrease once the γ parameter\ngets too high (2.59 in this figure). This is due to the fact that when penalty\nvalues of many possible edges get large quickly, their relative differences\nbecome less. Consequently, candidate paths become less. The great fluc-\ntuation of node 26 (Figure 7.10) is due to its position in the center of the\ntopology and equal distance to both gateway and destination. Similar ob-\nservations can be made about mutual information values of destination\nnode 16 (Figure 7.11).\nWe also observe from Figure 7.12 that our algorithm achieves our goal\nof preserving traffic pattern. In the first place, it is easy to conclude that in\n" }, { "page_number": 258, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n247\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\n0\n0.2\n0.4\n0.6\n0.8\n1\n1.2\n1.4\n1.6\n1.8\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\n(b) Observer: node 2\n(a) Observer: node 6\n0\n0.2\n0.4\n0.6\n0.8\n1\n1.2\n1.4\n1.6\n1.8\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\n0\n0.2\n0.4\n0.6\n0.8\n1\n1.2\n1.4\n1.6\n1.8\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\n(d) Observer: node 19\n(c) Observer: node 14\nFigure 7.11\nTraffic\nmutual information under different penalty parameters\n(destination: node 16).\nnormal shortest path routing, all relaying nodes share the same traffic infor-\nmation with the destination node, as shown by the tail of the ShortestPath\ncurve in Figure 7.12. However, for our algorithm, the mutual information\nshared between relaying nodes and the destination node varies much less\namong all relaying nodes, and the higher γ is, the more leveled off the curve\nbecomes and the closer we are to the goal of minimizing the greatest mu-\ntual information, formulated in Equation 7.6. It is also interesting to observe\nthat mutual information is 0 for some nodes far away from both gateway\nand destination; for example, in Figure 7.12(a), when destination is 1, while\nall nodes participate in relaying packets for destination 16, because desti-\nnation and gateway nodes are in opposite directions with respect to WMN\ntopology.\n7.6\nCollusion Analysis\nThe relative small size of a typical WMN makes it easy for spatially close\neavesdroppers to find each other. This alerts us to the high possibility of\ncollusion of two malicious observers by exchanging their observed traffic\n" }, { "page_number": 259, "text": "248\n■\nSecurity in Wireless Mesh Networks\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n5\n10\n15\n20\n25\n30\nTraffic entropy\nNode\nShortestpath\n1.12\n1.57\n1.85\n2.59\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n5\n10\n15\n20\n25\n30\nTraffic entropy\nNode\n(a) Destination: node 1\n(b) Destination: node 16\nShortestpath\n1.12\n1.57\n1.85\n2.59\nFigure 7.12\nSorted traffic mutual information under different penalty parameters.\npattern, and motivates us to make our proposed solution resilient to such\ncollusion threats.\nTo analyze the extent to which collusion reveals original traffic pattern,\nwe study the fluctuation of the observed traffic information. In this way,\nwe can know how much in addition the colluders can observe about the\noriginal traffic.\n7.6.1\nProblem Description\nPreviously, we focused on traffic confidentiality and studied the problem\nof traffic pattern concealment via routing control. However, the relative\nsmall size of a WMN, aided by the stationary adjacent routers, invites a\n" }, { "page_number": 260, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n249\n0.95\n1\n1.05\n1.1\n1.15\n1.2\n1.25\n1.3\n1.35\n1.4\n1.45\n1.5\n1.12\n1.31\n1.57\n1.85\n2.06\n2.59\nRatios\nParameter γ\nNode 1\nNode 6\nNode 11\nNode 15\nNode 17\nNode 24\nNode 29\nFigure 7.13\nCollusion reveals significant portion of original traffic pattern.\nhigh possibility of collusion of several observing relaying routers in the\ncommunity. Because it is highly possible that different observers will know\nabout various “ups and downs” of target’s traffic, if malicious observers\ninterchange their observed traffic information of target users, the combined\nobservation could reveal a significant portion of the original traffic pattern.\nThis is illustrated in Figure 7.13.\nGiven the size of the community network (less than 100 neighbor nodes),\nwe have a reasonable estimation that three or more malicious observers are\nunlikely to exist simultaneously, and hence we will focus on analyzing the\ncollusion problem of two observers in this work.\nThe parameters that affect significantly our collusion analysis include\nthe choice of cooperating observers and destination target node. Because\nany routing algorithm will largely depend on topology of the network, the\nrelative positions of observers and source and destination nodes can affect\nportions of revealed traffic pattern greatly. Another important parameter is\nthe base of the exponential penalty function explained in Section 7.4.\n7.6.2\nColluded Traffic Mutual Information\nOur modeling of colluded traffic analysis tries to study the influence of\ncollusion to observed traffic patterns of every period. This can help us to\nevaluate the resilience of our proposed A (PBSP) routing algorithm against\ncollusion attack. The notations used in this section are listed in Table 7.4.\n" }, { "page_number": 261, "text": "250\n■\nSecurity in Wireless Mesh Networks\nTable 7.4\nNotations Used in Section 7.6.2\nV\nWireless node set\nE\nEdge set\ng\nGateway node\nx\nDestination node\ny, z\nObserving nodes\nX\nRandom variable describing x’s traffic pattern\nY X, ZX\nRandom variables describing x’s traffic pattern\nobserved by y, z, separately\n(Y X, ZX)\nRandom variable describing x’s traffic pattern\nobserved by y, z together\nH(X)\nEntropy of X\nH(Y X)\nEntropy of Y X\nH(Y X, ZX, X)\nJoint entropy of Y X, ZX, X\nI (Y X; X)\nMutual information between X and Y X\nI (Y X, ZX; X)\nColluded mutual information between X and (Y X, ZX)\nIn what follows, we consider three nodes x and y, z. x is the destination\nnode of the traffic from the gateway g to x. Nodes y, z are the observing\nnodes, which relay packets for x and also try to analyze the traffic of x.\nDue to the uncertainty of routing, y, z may or may not be on the same\npath over time.\nTo begin with, we need to identify a measurement for colluded ob-\nservations. Based on the definition of traffic mutual information given in\nSection 7.3.2, we can measure the colluded observation about destination\nx with mutual information between x and (y, z). The traffic observations\nby y and z together can be deemed as the joint distribution of variable Y X\nand Z X. The colluded traffic mutual information I (Y X, Z X; X) of random\nvariable (Y X, Z X) with respect to X can then be defined as\nI (Y X, Z X; X) = H(Y X, Z X) + H(X) −H(Y X, Z X, X)\n(7.8)\nwhere H(Y X, Z X, X) is the joint entropy of Y X, Z X, and X. I (Y X, Z X; X)\ncan represent the information we could gain about X from (Y X, Z X), i.e.,\nfrom y, z together. Their relationship is shown in Figure 7.14.\n7.6.3\nSimulation Results\nFor ease of notation, in the following discussion, we would use H(Y, X)\nto denote H(Y X, X), i.e., the entropy of traffic that y observes about x.\nSimilarly, we simplify the joint traffic entropy H(Y X, Z X) as H(y, z, x),\nwhere Y X, Z X denote the portions of traffic that Y, Z observes about X. In\na subtly different way, we denote I (Y X; X) as I (Y; X) and I (Y X, Z X; X) as\nI (Y, Z; X).\n" }, { "page_number": 262, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n251\nH(Y x, Z x, X)\nI(Y x, Z x–, X)\nH(Zx)\nH(Y x)\nH(Y x, Z x)\nH(X)\nH(Z x)\nH(Y x)\nH(X)\nFigure 7.14\nVein graph representation of I (Y X, Z X; X), H(Y X, Z X), and H(Y X,\nZ X, X).\n7.6.3.1\nTraffic Curves\nIn the first place, we will present the measured traffic curves along a time\nline. In Figure 7.15, node 1 is the destination and we can easily conclude\nthat its traffic (node 1 observing itself ) is always the largest in amount.\nThis is because any node can observe the whole traffic of itself while other\nnodes can only observe a portion of it.\nAnother observation we can make is the fact that the colluded knowl-\nedge about traffic activity of node 1 (in squares), as expected, is higher\nthan any single observer, either 15 or 28. Moreover, we are confirmed by\nthis traffic curve figure that, although generally speaking, node 15 observes\nmuch more traffic of node 1; during some intervals, node 28 outperforms\n15 and elevates the aggregated knowledge about traffic activity of node 1.\nExample intervals are those near intervals 100 and 150.\n25\n20\n15\n10\n5\n0\n250\n200\n150\n100\n50\n0\nTraffic\nInterval\nTraffic(1)\nTraffic(15:1)\nTraffic(28:1)\nTraffic(15,28:1)\nFigure 7.15\nSampled traffic curves from experiment.\n" }, { "page_number": 263, "text": "252\n■\nSecurity in Wireless Mesh Networks\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n2\n4\n6\n8\n10\n12\n14\n16\n18\n20\nTraffic entropy\nNumber of periods\nShortestpath\n1.12\n1.57\n1.85\n2.59\nFigure 7.16\nColluded traffic mutual information (destination: 1, γ = 1.85).\n7.6.3.2\nColluded Traffic Mutual Information: Single Pair of Observers\nOur next results are the comparisons of colluded traffic mutual information\n(I (y, z; x)), single observer mutual information (I (y; x) and I (z; x)), original\ntraffic entropy (H(x)), separately observed traffic entropy (H(y, x) and\nH(z, x)), and joint entropy (H(y, z, x)).6 From our analysis in Section 7.6.2,\nwe can conclude the following relations among these values:\n1.\nH(y, x), H(z, x) ≤H(y, z, x) ≤H(x);\n2.\nI (y, x), I (z, x) ≤I (y, z, x) ≤H(x);\n3.\nI (y, x) ≤H(y, x) ≤H(x);\n4.\nI (z, x) ≤H(z, x) ≤H(x).\nNow we can verify if the simulation results shown in Figure 7.16\nsatisfy these relations. This means our modeling of traffic activity not only\ncharacterizes the traffic pattern fluctuation along the time, but also stands\nwith the test of collusion problem. The simulation results of our model\nconform with our conjecture.\nThe overlapping curves in Figure 7.16(b) indicate node 23 does not\nobserve any traffic of node 1. This could be true because 23 and 1 are on\nthe opposite side of the network.\nOn the other hand, Figure 7.17 shows similar results, except this time\nnode 16 is the destination.\n6 Please note that H(y, z, x), according to our notation, means H(Y X, Z X).\n" }, { "page_number": 264, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n253\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(16)\nH(6,16)\nH(28,16)\nH(6,28,16)\nI(6;16)\nI(28;16)\nI(6,28;16)\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(16)\nH(23,16)\nH(28,16)\nH(23,28,16)\nI(23;16)\nI(28;16)\nI(23,28;16)\n(a) Single pair of observers: 6, 28\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(16)\nH(21,16)\nH(29,16)\nH(21,29,16)\nI(21;16)\nI(29;16)\nI(21,29;16)\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(16)\nH(15,16)\nH(4,16)\nH(15,4,16)\nI(15;16)\nI(4;16)\nI(15,4;16)\n(c) Single pair of observers: 21, 29\n(b) Single pair of observers: 23, 28\n(d) Single pair of observers: 15, 4\nFigure 7.17\nColluded traffic mutual information (destination: 16, γ = 1.85).\n7.6.3.3\nColluded Traffic Mutual Information: Multiple Pairs\nof Observers\nNow that the simulation results have satisfied the necessary relations listed\nin the previous part, we would like to know how collusion can affect the\nperformance of the PBSP routing algorithm under discussion. To do so,\nwe will study the colluded traffic mutual information of several pairs of\nobservers in one figure. In this way, we can compare the ratio of traffic\ninformation revealing of different pairs of observers.\nFrom Figure 7.18 we can observe that the conditions above still hold.\nAdditionally, based on average values of the colluded traffic mutual infor-\nmation curves in both figures, we can guess that the PBSP algorithm still\nworks well when there are two observers colluding to share their knowl-\nedge about one destination.\nTo further confirm this conjecture, we can examine another set of simu-\nlation results, as shown in Figure 7.19. The colluded traffic mutual informa-\ntion of all observer pairs in this figure does not exceed half of total traffic\ninformation either. In Figure 7.19(b), however, we notice some small error\nof curves, i.e., the value of I (15, 6; 16) is a little less than that of I (15; 16)\n" }, { "page_number": 265, "text": "254\n■\nSecurity in Wireless Mesh Networks\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(1)\nI(21; 1)\nI(6; 1)\nI(28; 1)\nI(21, 6; 1)\nI(21, 28; 1)\nI(6, 28; 1)\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(16)\nI(21; 16)\nI(6; 16)\nI(28; 16)\nI(21, 6; 16)\nI(21, 28; 16)\nI(6, 28; 16)\n(a) Destination: 1, observers: 21, 6, 28\n(b) Destination: 16, observers: 21, 6, 28\nFigure 7.18\nColluded traffic mutual information (multiple pairs of observers,\nγ = 1.85).\nfor period 2. Although this is a small error, it reminds us of an approxima-\ntion when computing H(Y X, Z X, X). Instead of employing three parallel\nPacketCounters to get the aggregate traffic information, the simulation pro-\ngram approximates it based on the packet count value dictionary, which\nresults in a lower I (Y X, Z X; X) value.\nThe same explanation applies for the discrepancy in Figure 7.20(a). In\nthe meantime, the average value of colluded traffic mutual information of\nall observer pairs in Figure 7.20 remains approximately less than half of the\ntraffic entropy of the target node along the time.\n" }, { "page_number": 266, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n255\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(1)\nI(15; 1)\nI(6; 1)\nI(4; 1)\nI(15, 6; 1)\nI(15, 4; 1)\nI(6, 4; 1)\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(16)\nI(15; 16)\nI(6; 16)\nI(4; 16)\nI(15, 6; 16)\nI(15, 4; 16)\nI(6, 4; 16)\n(a) Destination: 1, observers: 15, 6, 4\n(b) Destination: 16, observers: 15, 6, 4\nFigure 7.19\nColluded traffic mutual information (multiple pairs of observers, γ =\n1.85).\n7.7\nRelated Work\nCurrently, multi-hop WMN is gaining more popularity, as deployments of\nWMN either serve as a substitute of traditional WLAN Internet connection,\nor aim at providing infrastructural large-scale network access [24].\nExisting research [3,7,10,19] on WMN has focused on how to better uti-\nlize the wireless channel resource and enhance its performance. For exam-\nple, some researchers [18] try to derive the optimal node density following\ncapacity analysis, while others strive to devise more efficient protocols [13].\nA survey paper by Akyildiz et al. [6] provides a good source for existing\n" }, { "page_number": 267, "text": "256\n■\nSecurity in Wireless Mesh Networks\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(1)\nI(21; 1)\nI(15; 1)\nI(7; 1)\nI(21, 15; 1)\nI(21, 7; 1)\nI(15, 7; 1)\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\nTraffic entropy\nPeriod\nH(16)\nI(21; 16)\nI(15; 16)\nI(7; 16)\nI(21, 15; 16)\nI(21, 7; 16)\nI(15, 7; 16)\n(a) Destination: 1, observers: 21, 15, 7\n(b) Destination: 16, observers: 21, 15, 7\nFigure 7.20\nColluded traffic mutual information (multiple pairs of observers, γ =\n1.85).\nand ongoing research about wireless mesh networks. Some of the pro-\nposed solutions include equipping mesh routers with multiple radios and\ndistributing the wireless backbone traffic over different wireless channels,\nrouting the traffic through different paths [15,33], or a joint solution of these\ntwo [25,26]. Theoretical study shows that these approaches can significantly\nincrease the capacity of WMN [21,22]. These results make a significant step\ntoward enabling WMN as an attractive alternative for broadband Internet\naccess.\nInformation theory is widely used and proves to be a useful tool. It\nworks in situations where variations are frequent and unpredictable and\n" }, { "page_number": 268, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n257\nhelps to identify pattern and extent of variation. Serjantov and Danezis [29]\ndefine an information theoretic anonymity metric and suggest developing\nmore sophisticated probabilistic anonymity metrics. Existing research [20]\nin the Internet setting employs information theoretical coding, which is\ntoo complex and impractical for WMNs. The book by David Mackay [23]\nprovides a good source for background knowledge in information theory.\nPrivacy has been a major concern of Internet users [12,31]. In the ex-\nisting literature of traffic pattern concealment, anonymous overlay routing\n[9,14,16,17,28,34] and traffic padding [30] have been proposed to preserve\nuser traffic privacy and increase the difficulty for traffic analysis [9,27]. The\nformer approach provides user anonymity in an end-to-end connection\nthrough layered encryption and multi-hop overlay routing. The latter one\nconceals the traffic shape by generating a continuous random data stream\nat the link level. However neither of them can be applied to WMN directly.\nFirst, the number of nodes in a WMN is limited. Second, the traffic forward-\ning relationship among nodes is strongly dependent on their locations and\nthe network topology. To better utilize the wireless channel resource and\nenhance the data delivery performance, a short path is usually selected\nor a load-balanced routing scheme is employed. Such observations show\nthat the anonymity systems, which rely on relaying traffic among nodes\n(randomly selected out of thousands) to gain anonymity, cannot effectively\npreserve users’ privacy in WMN, or do so at the cost of significant per-\nformance degradation. On the other hand, the traffic padding mechanism\nconsumes a considerable amount of network bandwidth, which makes it\nimpractical in resource-constrained WMNs.\nThe schemes designed in wireless ad hoc networks [11,32] are more\nfocused on location and identity privacy. While these are still issues in\nWMN, the traffic rates and temporal variations are more meaningful and\nconsequential.\nTo the best of our knowledge, no existing works have studied collusion\nproblems about traffic privacy in the scenario of wireless mesh networks.\n7.8\nConclusion\nThis chapter identifies the problem of traffic privacy preservation in wire-\nless mesh networks (WMN). To address this problem, we start by introduc-\ning a lightweight architecture for WMN, then propose “traffic entropy,” an\ninformation theoretic metric to quantify how well a solution performs at\npreserving the traffic pattern confidentiality, all of which pave the way to\nour penalty-based shortest path routing algorithm. Furthermore, we evalu-\nate our scheme against collusion of two malicious nodes. Simulation results\nshow that our algorithm is able to maximally preserve the traffic privacy,\nmeanwhile managing the network performance degradation within the\n" }, { "page_number": 269, "text": "258\n■\nSecurity in Wireless Mesh Networks\nacceptable region. Our simulation analysis also proves the resilience of\nour solution against two colluding observers.\nFor the future work, we will focus on the following problems. First,\nalthough our algorithm is evaluated in a single-radio, single-channel WMN\nsetting, it can be easily enhanced to exploit the advantage of multiple radios\nand multiple channels available in WMNs. Performance evaluation of the\nenhanced algorithm in such settings will be interesting. It is also beneficial\nto research the possibility of devising a distributed routing that achieves\nthe same goal, but supports better scalability.\nReferences\n[1]\nChaska wireless solutions. http://www.chaska.net/.\n[2]\nMesh Networks Inc. http://www.meshnetworks.com.\n[3]\nMIT Roofnet. http://www.pdos.lcs.mit.edu/roofnet/.\n[4]\nRadiant Networks. http://www.radiantnetworks.com.\n[5]\nSeattle Wireless. http://www.seattlewireless.net.\n[6]\nIan F. Akyildiz, Xudong Wang, and Weilin Wang. Wireless mesh networks:\nA survey. Comput. Netw. ISDN Syst., 47(4): 445–487, 2005.\n[7]\nMansoor Alicherry, Randeep Bhatia, and Li Li. Joint channel assignment and\nrouting for throughput optimization in multi-radio wireless mesh networks.\nIn Proceedings of ACM MOBICOM, 2005.\n[8]\nB. Awerbuch, D., Holmer, C. Nita-Rotaru, and H. Rubens. An on-demand\nsecure routing protocol resilient to byzantine failures. In ACM Workshop on\nWireless Security, 2002.\n[9]\nAdam Back, Ulf M¨oller, and Anton Stiglic. Traffic analysis attacks and trade-\noffs in anonymity providing systems. In Information Hiding Workshop (IH),\n2001.\n[10]\nJohn Bicket, Daniel Aguayo, Sanjit Biswas, and Robert Morris. Architecture\nand evaluation of an unplanned 802.11b mesh network. In Proceedings of\nACM MOBICOM, pp. 31–42, 2005.\n[11]\nS. Capkun, J.P. Hubaux, and M. Jakobsson. Secure and privacy-preserving\ncommunication in hybrid ad hoc networks. Technical report IC/2004/104,\nEPFL-DI-ICA, 2004.\n[12]\nRoger Clarke. Internet privacy concerns confirm the case for intervention.\nCommunications of the ACM, 42(2): 60–67, 1999.\n[13]\nDouglas S.J. De Couto, Daniel Aguayo, John Bicket, and Robert Morris.\nA high-throughput path metric for multi-hop wireless routing. In Proceed-\nings of ACM MobiCom, pp. 134–146, New York, ACM Press, 2003.\n[14]\nRoger Dingledine, Nick Mathewson, and Paul Syverson. Tor: The second-\ngeneration onion router. In USENIX Security Symposium, 2004.\n[15]\nR. Draves, J. Padhye, and B. Zill. Routing in multi-radio, multi-hop wireless\nmesh networks. In Proceedings of ACM MOBICOM, pages 114–128. ACM\nPress, 2004.\n[16]\nMichael J. Freedman and Robert Morris. Tarzan: A peer-to-peer anony-\nmizing network layer. In Proceedings of ACM CCS, 2002.\n" }, { "page_number": 270, "text": "Privacy Preservation in Wireless Mesh Networks\n■\n259\n[17]\nD. Goldschlag, M. Reed, and P. Syverson. Onion routing for anonymous\nand private internet connections. Communications of the ACM, 42(2):\n39–41, 1999.\n[18]\nP. Gupta and P. R. Kumar. The capacity of wireless networks. IEEE Trans-\nactions on Information Theory, 46(2): 388–404, 2000.\n[19]\nR. Karrer, A. Sabharwal, and E. Knightly. Enabling large-scale wireless\nbroadband: The case for taps. In HotNets, 2003.\n[20]\nSachin Katti, Dina Katabi, and Katarzyna Puchala. Slicing the onion: Anony-\nmous routing without PKI. MIT CSAIL Technical report 1000, 2005.\n[21]\nMurali Kodialam and Thyaga Nandagopal. Characterizing the capacity re-\ngion in multi-radio multi-channel wireless mesh networks. In Proceedings\nof ACM MOBICOM, 2005.\n[22]\nPradeep Kyasanur and Nitin H. Vaidya. Capacity of multi-channel wireless\nnetworks: Impact of number of channels and interfaces. In Proceedings of\nACM MOBICOM, pp. 43–57, New York, 2005.\n[23]\nDavid J.C. Mackay. Information theory, inference, and learning algorithms.\nCambridge, 2003.\n[24]\nKrishna Ramachandran, Milind M. Buddhikot, Scott Miller, Kevin Almeroth,\nand Elizabeth Belding-Royer. On the design and implementation of infras-\ntructure mesh networks. In Proceedings of IEEE WiMesh, 2005.\n[25]\nA. Raniwala and T. Chiueh. Architecture and algorithms for an IEEE\n802.11-based multi-channel wireless mesh network. In Proceedings of IEEE\nINFOCOM, 2005.\n[26]\nA. Raniwala, K. Gopalan, and T. Chiueh. Centralized channel assignment\nand routing algorithms for multi-channel wireless mesh networks. Mobile\nComputing and Communications Review, 8(2): 50–65, 2004.\n[27]\nJean-Franc¸ois Raymond. Traffic analysis: Protocols, attacks, design is-\nsues, and open problems. In International Workshop on Design Issues in\nAnonymity and Unobservability, 2000.\n[28]\nMichael G. Reed, Paul F. Syverson, and David Goldschlag. Anonymous\nconnections and onion routing. IEEE Journal on Selected Areas in Commu-\nnications, 16(4): 482–494, 1998.\n[29]\nA. Serjantov and G. Danezis. Towards an information theoretic metric for\nanonymity. In Proceedings of ACM MOBICOM, 2002.\n[30]\nW. Stallings. Cryptography and network security. Prentice Hall, 2003.\n[31]\nHuaiqing Wang, Matthew K.O. Lee, and Chen Wang. Consumer privacy\nconcerns about Internet marketing. Communications of the ACM, 41(3):\n63–70, 1998.\n[32]\nXiaoxin Wu and Bharat Bhargava. Ao2p: Ad hoc on-demand position-based\nprivate routing protocol. IEEE Transactions on Mobile Computing, 4(4):\n335–348, 2005.\n[33]\nYuan Yuan, Hao Yang, Starsky H.Y. Wong, Songwu Lu, and William\nArbaugh. Romer: Resilient opportunistic mesh routing for wireless mesh\nnetworks. In Proceedings of IEEE WiMesh, 2005.\n[34]\nLi Zhuang, Feng Zhou, Ben Y. Zhao, and Antony Rowstron. Cashmere:\nResilient anonymous routing. In Proceedings of USENIX NSDI, 2005.\n" }, { "page_number": 271, "text": "" }, { "page_number": 272, "text": "Chapter 8\nProviding Authentication,\nTrust, and Privacy in\nWireless Mesh Networks\nHassnaa Moustafa\nContents\n8.1\nIntroduction ......................................................... 262\n8.2\nSecurity Challenges in Wireless Mesh Networks ................... 263\n8.2.1\nMobility of Nodes ............................................ 263\n8.2.2\nHybrid Wireless Environment ............................... 264\n8.2.3\nCapacity and Density of Connections ....................... 264\n8.2.4\nIndividual Behavior of Nodes ............................... 265\n8.3\nThreats and Security Requirements in Wireless Mesh\nNetworks (WMNs)................................................... 266\n8.3.1\nThreats in WMNs Environment.............................. 266\n8.3.2\nDifferent Types of Attacks to WMNs ........................ 268\n8.3.3\nRequirements for Security Architectures and Mechanisms\nin WMNs ..................................................... 268\n8.4\nAuthentication ....................................................... 270\n8.4.1\n802.11i Authentication Model ............................... 272\n8.4.2\nData Packets Authentication................................. 275\n8.4.3\nAAA Architectures for WMNs................................ 276\n8.4.4\nExtensible Authentication Protocol Variants ................ 279\n8.4.4.1\nEAP with Token-Based Re-Authentication ........ 279\n261\n" }, { "page_number": 273, "text": "262\n■\nSecurity in Wireless Mesh Networks\n8.4.4.2\nEAP-TLS over PANA ................................ 280\n8.4.4.3\nEAP-TLS Using Proxy Chaining .................... 281\n8.4.5\nAAA in Multi-Operator WMNs............................... 282\n8.5\nTrust ................................................................. 283\n8.5.1\nUsing Reputation for Building Trust ........................ 284\n8.5.2\nDetecting Forwarding Misbehavior ......................... 285\n8.5.3\nTrusted Routing .............................................. 286\n8.6\nPrivacy ............................................................... 287\n8.6.1\nEfficient Key Distribution for Message Protection .......... 288\n8.6.2\nTraffic Privacy ................................................ 289\n8.6.3\nNon-Traceability ............................................. 290\n8.7\nConclusion and Outlook ............................................ 291\nReferences................................................................. 293\nSecurity is a big concern in wireless mesh networks (WMNs), where provid-\ning a robust secure system is considered one of the most critical challenges\npromoting the commercial deployment of WMNs and influencing their us-\nage. The security requirements in WMNs will determine what type of link\nlevel security protection is needed, at what protocol level intrusion detec-\ntion and prevention must be performed, and what amount of overhead due\nto security can be tolerated in the network. This will be a constant battle\nrequiring continuous security enhancements, continuous monitoring, and\nrapid responses to intrusions. This chapter starts by discussing the security\nchallenges in WMNs, showing the possible types of attacks in these net-\nworks, and stating the different security requirements. Then the problem of\nauthentication is presented, showing some authentication mechanisms that\nare useful in WMNs. The different contributions, employing the emerging\nstandards for authentication and secure links setup with a mobility manage-\nment support are presented, and the role of authentication, authorization,\nand accounting (AAA) in such environment is illustrated. The importance of\ntrust provision is shown, where security mechanisms will have to leverage\nspecial capabilities to detect untrusted elements and to protect the mesh’s\nintegrity. The chapter ends by discussing privacy provision in WMNs con-\nsidering traffic privacy and confidential transfer.\n8.1\nIntroduction\nWireless mesh networks (WMNs) have emerged as a key technology for\nnext-generation wireless networks, showing rapid progress and inspiring\nnumerous applications. WMNs, however, are not yet ready for wide-scale\ndeployment due to two main reasons: the interference caused by the wire-\nless communication and the non-security guarantees. The fact that all wire-\nless communications are prone to interference causes delay constraints\n" }, { "page_number": 274, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n263\nin WMNs. Nevertheless, it is believed that technological solutions would\nbe able to overcome this problem, for example, using multi-radio and\nmulti-channel Terminal Access Points (TAPs) [1]. The lack of security guar-\nantees is another factor slowing down the deployment of WMNs. In fact,\nsecurity in WMNs is still in its infancy and very little attention has been de-\nvoted thus far to this topic by the research community. As these networks\ncontinue to grow and as access to the mesh is available for any wireless-\nenabled device, it should be ensured that only authorized users are granted\nnetwork access. There is still a strong need for efficient solutions adapted\nfor different security requirements and for different usage scenarios. These\nsolutions have to counter attacks in all protocol layers, guaranteeing collab-\norative behaviors between mobile nodes. Trust relationships should exist\namong stakeholders for authentication, authorization, and accounting of\nend users. Well-performing tools need to be developed for mesh design,\nmaintenance, and management such that future mesh networks should be\nself-managed rather than unmanaged ones [2]. A number of challenges have\nto be considered during the design of security mechanisms and solutions,\nand appropriate security requirements should be defined considering the\ndifferent existing threats.\n8.2\nSecurity Challenges in Wireless Mesh Networks\nWMNs have special characteristics distinguishing them from other network\ntechnologies and consequently imposing a broad range of design chal-\nlenges to be solved. This section gives an overview on various security\nchallenges and requirements in WMNs. Security in WMNs is one of the\nwidely discussed topics and one of the major inherent caveats of wireless\nad hoc networking. Classical security approaches suffer from the inade-\nquate usage of redundant paths, and hence could not be directly applied\nin WMNs. The mobility of nodes, the hybrid wireless environment created\nby the different wireless mesh architectures, the density of connections in\nthese networks, and the unpredictable behavior of nodes are critical factors\ninfluencing the security requirements of WMNs and posing new security\nchallenges. One possible approach in providing practically feasible solu-\ntions is to deploy, combine, and adopt existing security approaches and\nprotocols in wireless networks in general, and in ad hoc networks in par-\nticular. However, specific security mechanisms must be developed allowing\nintense load sharing while taking into account local capacity limitations and\ndynamic load changes.\n8.2.1\nMobility of Nodes\nAn attractive point in commercial WMN deployment is the seamless access\nof mobile clients to services offered by these networks, in a completely\n" }, { "page_number": 275, "text": "264\n■\nSecurity in Wireless Mesh Networks\ntransparent manner to clients’ mobility. However, clients’ mobility itself is\na challenge which poses some constraints on WMN security. A part of this\nchallenge lies in the mobile devices themselves. First, mobile devices are\nsusceptible to thefts and thus can be misused by attackers in either an\nunauthorized access or a communication corruption. Second, the fact that\nmost mobile devices are “thin clients” of limited power, CPU, and storage\ncapacity, leads to difficulty in running some security mechanisms in WMN\n(as for example, encryption algorithms requiring special resources).\nAnother part of this challenge arises from the mobility itself, where\nmobile clients are susceptible to roaming across different administrative\ndomains that may have different security policies. Thus, efficient security\nmechanisms are needed for handling clients’ roaming in a secure manner.\nFinally, the mobility of mobile clients can facilitate tracing the mobile clients’\nexistence at different places. Privacy protection mechanisms are thus impor-\ntant so that an attacker could not hack client privacy by tracing its mobility.\n8.2.2\nHybrid Wireless Environment\nWMNs are expected to offer seamless wireless network access for mobile\nusers within a hybrid wireless environment. In such an environment, hybrid\nwireless communication allows multi-hop access mode combining peer-to-\npeer communication between mobile nodes as well as mobile nodes’ com-\nmunication with a fixed infrastructure. Peer-to-peer communication can be\nconsidered as pure ad hoc networks’ communication. In addition, each mo-\nbile node may access a fixed infrastructure either directly or via other nodes\n(mesh routers) in a multi-hop fashion. In spite of the seamless access feature\nprovided by WMNs, there are no mechanisms in place implementing se-\ncurity services when a mobile terminal roams between disparate networks.\nConsequently, some essential features like secure roaming, authentication,\nand authorization should be highly considered in that type of environment.\nThe security mechanisms must guarantee that only authorized users can use\nthe network resources and access the services offered by the provider. Fur-\nthermore, eavesdropping as well as the modification of the transmitted data\nduring the multi-hop communication, must be prevented. There is a lack of\nefficient security mechanisms that offer secure links setup and confidential\ndata transfer among mobile clients in hybrid wireless environments. This is\nin part because the security and mobility management solutions, in wire-\nless networks in general, are often implemented at different protocol layers\nwith limited amount of interaction between these layers.\n8.2.3\nCapacity and Density of Connections\nThe capacity of WMNs is an important issue that is worth consideration\nduring the development of security mechanisms in these networks. Many\n" }, { "page_number": 276, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n265\nfactors can affect the capacity of WMNs such as network architecture, mo-\nbile nodes’ density, number of channels used for each mobile node, trans-\nmission power level, and nodes’ mobility [3]. Hence, a clear understanding\nof the relationship between network capacity and the above factors pro-\nvides guidelines for protocols’ development as well as architecture design\nin WMNs. Nevertheless, the current security mechanisms and protocols in\nWMNs do not take this fact into account, although some security issues are\nrelated to radio resource’s management or can arise due to the nature of\nthe radio medium and the resource constrained devices. Consequently, a\nnumber of problems take place due to non-coherence between the security\nmechanisms and the capacity and density of connections in WMNs. Some\nexamples:\n■\nThe possibility of spoofing power control messages among nodes,\nwhich can result in an unstable situation within a group of mesh\ncells, causing loss of services and increasing the load in the neigh-\nboring mesh cells.\n■\nThe power resources constraint poses an obstacle to running key\nmanagement protocols in high-density mesh cells, requiring a lot of\nmessages and keys exchanges.\n■\nThe difficulty in managing cryptography over all the mesh connec-\ntions, especially in mesh cells of high connection density.\n■\nThe decentralized authentication process, which is a significant re-\nquirement in WMNs, becomes more complex in high-density mesh\ncells, and adequate authenticators’ delegation should take place.\n8.2.4\nIndividual Behavior of Nodes\nCooperation among nodes is a primary requirement for WMN functioning.\nNode cooperation in WMNs is critical for multi-hop transmission, collective\ndata processing, and cooperative security functions. However, providing\nservice to each other consumes resources, which are generally scarce in\nmobile nodes. Thus, cooperation cannot be taken for granted, especially\nin opened mesh networks scenarios, because each user would prefer to\nmaximize his own benefit while minimizing his contribution. Mobile nodes\nin WMNs are supposed to be rational in the sense that they try to maximize\ntheir own utilities in a self-interested way. The cooperation issue concerns\ndifferent layers of the node’s protocol stack, with different aims and ways\nof acting, where a self-interested node can misbehave by:\n1.\nNon-adherence to the protocols specification\n2.\nOptimization of a particular utility function, possibly at the expense\nof other nodes\n" }, { "page_number": 277, "text": "266\n■\nSecurity in Wireless Mesh Networks\nConsequently, selfishness and greediness are two misbehaviors that are\nlikely to take place in WMNs. Nodes may behave selfishly by not forward-\ning packets for others to save power, bandwidth, or just because of secu-\nrity and privacy concerns. Watchdog [4], Confidant [5], and Catch [6] are\nthree approaches developed to detect selfishness and enforce distributed\ncooperation and are suitable for WMNs. Watchdog is based on monitoring\nneighbors to identify a misbehaving node that does not cooperate during\ndata transmission. However, Confidant and Catch incorporate an additional\npunishment mechanism making misbehavior unattractive through isolating\nmisbehaving nodes. On the other hand, a node may behave greedily in\nconsuming channel and bandwidth for its own benefits at the expense of\nthe other users. A mechanism that modifies 802.11 for facilitating the de-\ntection of greedy nodes is proposed in [7]; also the DOMINO mechanism\n[8] solves the greedy sender problem in 802.11 WLANs with a possible\nextension to multi-hop wireless networks and WMNs.\nTo provide secure cooperation mechanisms that are suitable for WMNs,\nthe following factors are important to be considered:\n■\nThe vulnerability of wireless links, compared to wired ones, in terms\nof eavesdropping and jamming\n■\nThe weak connection of each node with the network authority\n■\nThe fact that devices are becoming more and more programmable\nBecause the above mechanisms require maintaining a great deal of state\ninformation at each node while monitoring its neighbors, adaptive schemes\nare needed for right functioning in WMNs. Two other important issues to\nbe considered are the distributed detection of selfishness and greediness\nmisbehaviors and providing incentives to mobile nodes to stimulate coop-\neration.\n8.3\nThreats and Security Requirements in Wireless\nMesh Networks (WMNs)\nBecause WMNs are based on the concept of wireless distribution system\n(WDS), they are vulnerable to a variety of threats. Security measures should\nbe taken to avoid these threats and allow reliable communication. Also, the\nnotion of WDS requires end-to-end security assurance for each end user.\n8.3.1\nThreats in Wireless Mesh Network Environment\nThreats in WMNs are mainly due to the nature of the radio links, the ubiquity\nof wireless communications, and the multi-hop communication. The main\n" }, { "page_number": 278, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n267\ntarget of security solutions in WMNs is to encounter the following types of\nsecurity threats [9]:\n■\nEavesdropping and data modification: The nature of radio environ-\nment can cause eavesdropping and modification of the data sent\nby mobile nodes. The presence of wireless links and intermediate\nmobile nodes in WMNs requires the existence of encryption and\nintegrity protection mechanisms to prevent eavesdropping and data\nmodification, allowing confidentiality and integrity of the transmitted\ndata.\n■\nUnauthorized access: The possibility of setting up wireless connec-\ntions to any mesh network can result in unauthorized nodes access\nto WMNs, posing a critical threat to these networks. In closed WMNs,\nwhich have a centralized administration, successful authentication\nshould be a requirement for joining a mesh network. However, in\nopen mesh networks with no central control, alternative solutions\nmust be in place to allow authentication between mobile nodes in\na distributed manner.\n■\nDenial of service (DoS): A traditional DoS may take place during\nmulti-hop transmission by an intermediate mobile node selectively\ndropping traffic frames. The DoS characteristic of WMNs is gener-\nally caused by routing misbehavior of a mobile node. The black hole\nattack [10] is an example of the DoS, where the malicious mobile\nnode can tamper with the routing messages in a network, or spoof\nthe MAC address of a mobile node into claiming a fake shortest\npath so as to get all the packets routed to itself, without any inten-\ntion to route the packets to destination. Indeed, any mobile node\nthat is correctly authenticated when joining the mesh network may\nsuddenly start misbehaving causing DoS. Thus, it is very difficult to\ndiscover and prevent the DoS in WMNs.\nCountermeasures need to be devised for WMNs using the security op-\ntions according to the size of risks. An intrusion detection system may be\nused in such case to address some of the threats. A useful approach to\ncounter security threats is to study the threats with respect to their likeli-\nhood of occurrence, their possible impact on individual users and on the\nwhole system, and the expected risk from these threats [11]. The likelihood\nevaluates the possibility of conducting attacks related with the threat, tak-\ning into account the motivation for an attacker and the technical difficulties\nthat he needs to resolve. The impact can evaluate the consequences of an\nattack related to the threat. This depends on whether the attack is directed\nto an individual user or to the whole system. It also depends on the pos-\nsibility of service loss caused by the attack. Consequently, the risk can be\ndefined as a function of the likelihood and the impact values.\n" }, { "page_number": 279, "text": "268\n■\nSecurity in Wireless Mesh Networks\n8.3.2\nDifferent Types of Attacks to WMNs\nAttacks can exist at different layers in WMNs causing network failure. At the\nphysical layer, an attacker may jam the transmission of wireless antennas or\nsimply destroy the hardware of a certain node. At the MAC layer, an attacker\nmay abuse the fairness of medium access by sending MAC control and\ndata packets or impersonating a legal node. Attacks may occur in routing\nprotocols such as advertising wrong routing updates. At the application\nlayer, an attacker could inject false fake information, thus undermining\nthe integrity of the application. Attackers may also sneak into the network\nby misusing the cryptographic primitives. Consequently, the exchange of\ncryptographic information should take place through special schemes, for\nexample, the rational exchange scheme [12], ensuring that a misbehaving\nparty cannot gain anything from misbehavior. Furthermore, the absence of\na central authority, a trusted third party, or a server to manage security keys\nnecessitates distributed key management.\nTwo classes of attacks are likely to occur in WMNs:\n1.\nExternal attacks, in which attackers not belonging to the network\njam the communication or inject erroneous information, mostly take\nplace at open mesh networks that are not controlled by a central\nauthority.\n2.\nInternal attacks, in which attackers are internal, compromised nodes\nthat are difficult to be detected.\nBoth types of attacks may be either passive (intending to steal information\nand to eavesdrop on the communication within the network) or active\n(modifying and injecting packets to the network).\nGenerally, there are two approaches to dealing with security attacks:\nprevention and detection. Prevention aims at thwarting security breaches\nfrom occurring in the first place, whereas detection and reaction are nec-\nessary in case of prevention failure. On the other hand, detection aims at\ndiscovering malicious nodes that carry out attacks to the network. Special\nmechanisms can be in place to detect attackers, for example, intrusion de-\ntection mechanisms. However, it is difficult to detect internal attackers even\nin the presence of detection mechanisms. The ideal method is integrating\nthe two approaches; however, the cost of a security system in this case may\nbe too expensive for mobile nodes in this environment. We notice that most\nof the security mechanisms and protocols follow the prevention approach.\n8.3.3\nRequirements for Security Architectures\nand Mechanisms in WMNs\nThe existence of robust authentication mechanisms is an important secu-\nrity requirement in WMNs to prevent unauthorized user access. Mutual\n" }, { "page_number": 280, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n269\nauthentication of mesh nodes is a critical issue that should be satisfied. It is\nimportant to distinguish between the nodes’ authentication at the initializa-\ntion phase and the nodes’ authentication during the session while sending\nand receiving packets. For authenticating mobile nodes at the initializa-\ntion phase, public key cryptography can be useful in closed mesh network\nscenarios. Mutual authentication can take place in this case through using\ncertified public/private key pairs assigned to the mobile nodes by the op-\nerator that is managing them. However, the use of public key cryptography\nto authenticate mobile nodes during the session is a heavy process causing\nimportant delay constraints. Instead, the nodes can rely on symmetric key\ncryptography, using session keys which they establish during the initializa-\ntion phase or long-term shared keys that can be originally loaded in the\ndevices. In open mesh network scenarios, using per-session per-connection\nkeys seems a feasible solution while considering the knowledge of the key\nas a stepping stone for authentication.\nOnce the nodes are authenticated, it is necessary to ensure the integrity\nof the exchanged messages and prevent messages modification. A possible\nway to do so is through using symmetric keys that are derived during the\nsession establishment. Consequently, employing encryption mechanisms in\nWMNs can assure the integrity and confidentiality of transmissions, where\nreliable encryption solutions are needed while minimizing complexity and\noverhead. These solutions should allow hop-by-hop encryption and should\navoid the possibility of eavesdropping on or tampering with the data by\nintermediate mobile nodes.\nHybrid security architectures are mostly suitable in WMNs, comprising\ntwo phases. The first phase concerns mutual authentication and encryp-\ntion [13]. In mutual authentication phase, a public key infrastructure (PKI)\nis generally applied. However, this step requires the deployment of a cen-\ntral node functioning as a trust center and running a database against which\nkey verification can take place. The authenticity of central nodes can also\nbe verified by public/private keying. Based on this authentication, the sec-\nond step is the exchange of symmetric keys per connection to encrypt all\ndata transfer. This second step can be optional, because the mutual authen-\ntication enables a security level that can be sufficient for many systems. On\nthe other hand, the encryption can pose relatively high requirements on\nthe node’s resources.\nConsidering the characteristics of WMNs, security mechanisms and pro-\ntocols should satisfy most of the following requirements:\n■\nScalability: The performance of protocols and mechanisms, in terms\nof computational and communication cost, should not degrade with\nthe network size. To achieve this, every node should not be required\nto have the global knowledge of the network, for example, sharing\na pairwise key with every other node in the network.\n" }, { "page_number": 281, "text": "270\n■\nSecurity in Wireless Mesh Networks\n■\nEfficiency: Mechanisms and protocols must be resource efficient.\nAlthough security should have a cost, the protocols should incur\nas little overhead as possible. Security mechanisms and protocols\nshould not require large bandwidth overhead and operations that\nrequire high computations such as those based on public key tech-\nniques should be minimized.\n■\nRouting protocol independence: One important point in designing\nsecurity mechanisms and protocols is the independence of the rout-\ning protocol. Although it is possible to design mechanisms that work\nwith specific routing protocols, this would require the design of a\nnew customized protocol for every routing protocol, which is clearly\nundesirable.\n■\nTransparency: It is undesirable that the deployment of security mech-\nanisms requires modification or redesign of other protocols in the\nprotocol stack. Security mechanisms and protocols should work\ntransparently with other protocols and without affecting the func-\ntionality of other protocols such as routing protocols or application\nlayer protocols.\n■\nFast authentication: There should be no high delay for authentica-\ntion. Otherwise, the authentication latency would be unacceptably\nhigh in such a multi-hop communication environment, especially\nwhen authentication is needed between different administrative\ndomains.\n8.4\nAuthentication\nAuthentication of mobile nodes in WMNs can assure authorized clients par-\nticipation. The simplest solution is to employ an authentication key shared\nby all nodes in the network. Although this mechanism is simple, it has the\nfollowing disadvantages:\n■\nAn attacker only needs to compromise one node to break the secu-\nrity of the system and paralyze the entire network.\n■\nIf the global key is divulged, it is not possible to identify the com-\npromised node.\n■\nIt is expensive to recover from a compromise as it usually involves\na group key update process.\n■\nMobile nodes do not usually belong to the same community, which\nleads to a difficulty in installing/pre-configuring the shared keys.\nAnother well-known approach that can provide strong source authen-\ntication is attaching digital signature to packets. However, signing every\npacket can be prohibitively expensive because the computational capacity\nand battery power of mobile nodes are quite constrained. Therefore, the\n" }, { "page_number": 282, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n271\nchallenge is to design authentication mechanisms for the more vulnerable\nyet more resource-constrained environment of WMNs.\nAuthentication and authorization are important counter-attack measures\nin WMN deployment, allowing only authorized users to get connections via\nthe mesh network and preventing adversaries to sneak into the network\ndisrupting the normal operation or service provision. Authentication, au-\nthorization, and accounting (AAA) are provided in most of the WLANs\napplications and commercial services through a centralized server such as\nRADIUS or DIAMETER. However, the centralized scheme is not appropriate\nin the case of multi-hop WMNs and secure key management is much more\ndifficult. Thus, distributed authentication and authorization schemes with\nsecure key management are required in such an environment. Because\nWMNs can be managed by more than one operator/provider, authentica-\ntion should be performed during mobile nodes’ roaming across different\nwireless mesh routers and across different administrative domains. This al-\nlows users’ mobility with seamless and secure access to the offered services\nin the mesh network. A possible approach for distributed authentication is\nthe continuous discovery and mutual authentication between neighbors,\nwhether they are mobile clients or fixed/mobile mesh nodes. Nevertheless,\nif mobile nodes move back to the range of previous authenticated neigh-\nbors or mesh nodes, it is necessary to perform re-authentication to prevent\nan adversary from taking advantage of the gap between the last association\nand the current association with the old neighbor to launch an imperson-\nation attack. The IEEE 802.11i standard proposed the storage of session\nkeys at authenticators to mitigate the overhead of re-authentication; how-\never, it is vulnerable to impersonation attacks, in which a malicious access\npoint can use previously stored keys to dupe user nodes. Other vendors’\nspecific solutions are proposed by Cisco, Aruba, and Trapeze networks,\nintegrating a switched architecture in the 802.11i authentication aiming to\ncentralize the storage of the authentication keys, therefore to accelerate\nthe re-authentication. These solutions work well in WLAN applications, re-\nsolving expensive overhead of re-authentication. However, there are no\nassociated security mechanisms to prevent attacks on stored keys. As well,\nthese solutions are not scalable to WMNs, where decentralized key man-\nagement is necessary.\nThe following sub-sections describe some authentication mechanisms\nand protocols that are useful for application in WMNs. Four approaches\nare mainly considered:\n1.\nAdapting the 802.11i authentication to the mesh network environ-\nment to authenticate nodes and to allow secure links setup at layer 2.\n2.\nAuthenticating data packets transmitted or received aiming to pre-\nvent non-authorized nodes from injecting erroneous packets in the\nnetwork.\n" }, { "page_number": 283, "text": "272\n■\nSecurity in Wireless Mesh Networks\n3.\nUsing new AAA infrastructures adapted to the dynamic and decen-\ntralized WMN environment.\n4.\nExtending some existing authentication protocols to the WMN en-\nvironment.\n8.4.1\n802.11i Authentication Model\nIn most commercial deployments of WLANs, IEEE 802.11i [14] is the most\ncommon approach for assuring authentication and secure links setup at\nlayer 2. However, the IEEE 802.11i authentication does not fully address\nthe problem of WLAN vulnerability. In IEEE 802.11i authentication, as de-\npicted in Figure 8.1, the mobile station and the authentication server (AS)\napply the 802.1X [15] authentication model carrying out some negotiation\nto agree on Pairwise Master Key (PMK) by using some upper layer au-\nthentication schemes or using a pre-shared secret. This key is generated by\nboth the mobile client and the AS, assuring the mutual authentication be-\ntween them. The access point (AP) then receives a PMK copy from the AS,\nauthenticating the mobile client and authorizing its communication. After-\nward, a four-way handshake starts between the AP and the mobile station\nto generate encryption keys from the generated PMK. Encryption keys can\nassure confidential transfer between the mobile station and the AP. If the\nmobile station roams to a new AP, this mobile station will perform another\nfull 802.1X authentication with the AS to derive a new PMK. For perfor-\nmance reasons, the PMK of the mobile station can be cached by the mobile\nstation and the AP to be used for later re-association without another full\nRadius (EAP-TLS)\nEAPoL/EAP-\nTLS\nAssociation\nphase\nMN\nAP (authenticator)/\nradius client\nAS/radius server\nMN\nAuthenticator\nPMK\nPMK\nPMK\nPTK\nPTK\nPMK exists\nMutual authentication phase\n4-Way hardshake\n4-Way hardshake (EAPoL-Key messages)\nto generate cryptographic keys through\nfirst generating the pairwise transient\nkey (PTK)\n \nFigure 8.1\nIEEE 802.11i authentication model.\n" }, { "page_number": 284, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n273\nauthentication. The features of 802.11i exhibit a potential vulnerability be-\ncause a compromised AP can still authenticate itself to a mobile station and\ngain control over the connection. Furthermore, IEEE 802.11i authentication\ndoes not provide a solution for multi-hop communication. Consequently,\nnew mechanisms are needed for authentication and secure layer 2 links\nsetup in WMNs.\nWireless Dual Authentication Protocol (WDAP) [16] is proposed for\n802.11 WLAN and can be extended to WMNs. WDAP provides authenti-\ncation for both mobile stations and access points and overcomes the short-\ncomings of other proposed mutual authentication protocols. The name\n“dual” returns to the fact that the AS authenticates both the mobile sta-\ntion and access points. As in the four-way handshake in IEEE 802.11i, this\nprotocol also generates a session key for confidentiality of communica-\ntions between the mobile station and the AP after a successful authenti-\ncation. WDAP provides authentication during the initial connection state\nand while roaming including three sub-protocols: an authentication proto-\ncol, a de-authentication protocol, and a roaming authentication protocol.\nFigure 8.2 illustrates the WDAP authentication process. In the authentica-\ntion protocol, the AP that receives the mobile station authentication request,\ncreates also an authentication request for itself concatenating this request\nto the received request from the mobile station and sending the concate-\nnated request to the AS. The dual part of WDAP lies in this phase, because\nboth the mobile station and the AP do not trust each other until the AS\nauthenticates both of them. In case of successful authentication, a session\nAssociation phase\nMN authentication request\nConcatenated authentication request\n(MN request + AP request)\nAS\nAP (authenticator)\nMN\nChallenged request/response\nAuthentication success\nCopy of session key\nSession key\nEncrypted copy of session key\nFigure 8.2\nAuthentication in WDAP.\n" }, { "page_number": 285, "text": "274\n■\nSecurity in Wireless Mesh Networks\nkey authenticating both the AP and the mobile station is generated by the\nAS and sent to the AP. The AP then sends this key to the mobile station\nencrypting it with the mobile station secret key. This key is thus shared\nbetween the AP and the mobile station for their secure communication\nand secure de-authentication when the session is finished. When a mobile\nstation finishes a session with an AP, secure de-authentication takes place\nto prevent the connection from being exploited by an adversary. In case\nof a mobile station roaming to a new AP, it sends out a roaming authen-\ntication request message to the new AP, where the new AP concatenates\nits authentication request to this message and then sends the concatenated\nrequest to the AS. After the AS verification of the previous authentication of\nthe mobile station and the successful authentication of the new AP, it sends\na session key revoke message to the old AP and a new generated session\nkey to the new AP to be shared with the mobile station. Applying WDAP\nin WMN environments allows the mutual authentication between mobile\nnodes and WMRs. Also, WDAP can be used to assure the authentication\nbetween the WMRs themselves through authentication requests concatena-\ntion. In case of multi-hop communication in WMNs, each pair of nodes can\nmutually authenticate through the session key generated by the AS. How-\never, a solution is needed in case of open mesh networks scenarios, where\nthe AS is not always in place. Another problem comes from the roaming\nauthentication approach in WDAP which is not quite suitable for WMN\nenvironments, as it restricts the roaming to only new APs and does not\nconsider the case of “back roaming” where the mobile node might need to\nre-connect with another mobile node or an AP with whom it was authen-\nticated before. Consequently, the WDAP session key revoke mechanism\nbrings some disadvantages to WMNs and another mechanism is required.\nAn approach that adapts IEEE 802.11i to the multi-hop communication\nis presented in [17]. An extended forwarding capability in 802.11i is pro-\nposed without compromising its secure features, to set up authenticated\nlinks on layer 2 and achieve secure wireless access as well as confidential\ndata transfer in ad hoc multi-hop environments. The general objective of\nthis approach is supporting mobile clients’ secure and seamless access to\nthe Internet, near public WLAN hotspots, even when they move beyond\nWLAN communication ranges. To accomplish the AAA process for a mo-\nbile client existing in the WLAN communication range, classical 802.11i\nauthentication and messages’ exchange takes place. On the other hand,\nas illustrated in Figure 8.3, for accomplishing the AAA process for mo-\nbile clients that do not exist in the WLAN communication range and are\nconsequently belonging to ad hoc clusters, 802.11i is extended to support\nforwarding capabilities. In this case, the notion of friend nodes is intro-\nduced allowing each mobile client to initiate the authentication process\nthrough a selected node in its proximity. The friend node plays the role of\nan auxiliary authenticator and forwards the authentication request of the\n" }, { "page_number": 286, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n275\nMN\nFriend node (auxiliary authenticator)\n/radius client\nAuxiliary authenticator\n/radius proxy\nAP (authenticator)/\nradius proxy\nAS/radius\nserver\nAssociation\nphase\nRadius\n(EAP-TLS)\nRadius (EAP-TLS)\nPMK\nPMK\nPMK\nPMK\nPMK\nEAPoL/EAP-\nTLS\nFigure 8.3\nAdapted 802.11i with EAP-TLS for multi-hop communication.\nmobile node to the actual authenticator (which is the AP in this case). If the\nfriend node does not fall in the communication range of the AP, it invokes\nother friend nodes in a recursive manner until reaching the AP. The con-\ncept of proxy RADIUS [18] is used for forwarding compatibility and secure\nmulti-hop messages’ exchange, where proxy chaining [19] takes place if the\nfriend node is not directly connected to an AP. To obtain increased security\non each authenticated link between each communicating parties, 802.11i\nencryption phase takes place through employing the four-way handshake\nbetween each mobile node and its authenticator (AP or friend node). This\napproach is useful in open mesh network scenarios to allow authentica-\ntion by delegation among mesh nodes. In addition, this approach allows\nauthentication keys storage among intermediate nodes, which optimizes\nthe re-authentication process in case of mobile nodes’ roaming. However,\nan adaptation is needed in terms of allowing multiple connections to au-\nthenticators whether APs or auxiliary authenticators (friend nodes) in case\nof a dense mesh topology. Also, a solution is needed to support fast and se-\ncure roaming across multiple WMRs. A possible solution is through sharing\nsession keys of authenticated clients among WMRs.\n8.4.2\nData Packets Authentication\nAuthenticating transmitted data packets is another approach preventing\nunauthorized nodes’ connection to the WMNs. A Lightweight Hop-by-hop\n" }, { "page_number": 287, "text": "276\n■\nSecurity in Wireless Mesh Networks\nAccess Protocol (LHAP) [20,21] is proposed for authenticating mobile clients\nin wireless dynamic environments, preventing resource consumption at-\ntacks through employing packet authentication. LHAP implements light-\nweight hop-by-hop authentication, where intermediate nodes authenticate\nall the packets they receive before forwarding them. This protocol allows a\nmobile node to first perform some inexpensive authentication operations to\nbootstrap a trust relationship with its neighbors, then to apply a lightweight\nprotocol for subsequent traffic authentication. LHAP is mainly proposed for\nad hoc networks, where it resides between the data link layer and the net-\nwork layer and can be seamlessly integrated with secure routing protocols\nto provide a more secure ad hoc network.\nLHAP employs a packet authentication technique based on the use of\none-way hash chains [22]. Also, LHAP uses Tesla [23] to reduce the number\nof public key operations for bootstrapping and maintaining trust between\nnodes. For every traffic packet received from the network layer, LHAP adds\nits own header, which includes its node ID, a packet type field indicating a\ntraffic packet, and an authentication tag. Afterward, LHAP passes the packet\nto the data link layer and generates its own control packets for establishing\nand maintaining trust relationships with neighbor nodes. For a received\ntraffic packet, LHAP verifies its authenticity based on the authentication\ntag in the packet header. If the packet is valid, LHAP removes the LHAP\nheader and passes the packet to the network layer; otherwise, it discards\nthe packet. LHAP control packets are not passed to the network layer with\nthe goal to allow LHAP execution without affecting the operation of other\nprotocols’ layers.\nThis protocol is quite adaptable to WMN environments, especially open\nmesh scenarios when the AS is not in place, preventing unauthorized\nclients’ participation in the communication and allowing hop-by-hop au-\nthentication. For secure roaming, LHAP can be useful in distributing session\nkeys among mobile clients employing a special type of packet designated\nfor this issue. However, the focus of this protocol on resource consumption\nattacks’ prevention restricts its application to a number of scenarios. Also,\nthe fact that LHAP does not prevent insider attackers from carrying out ma-\nlicious actions necessitates complementary solutions with such protocol.\n8.4.3\nAAA Architectures for WMNs\nWMN deployment requires appropriate architectures for the different\ntypes of scenarios. An important step toward the wide commercial deploy-\nments of WMNs is the trust relationship between stakeholders of different\naccess networks, each having its own security mechanisms. To provide\nseamless service across heterogeneous access networks, there must be a\ntrust relationship among the stakeholders for authentication, authorization,\nand accounting, and billing of end users.\n" }, { "page_number": 288, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n277\nA lightweight AAA infrastructure is proposed in [24] providing continu-\nous, on-demand, end-to-end security in heterogeneous networks including\nWMN scenarios. This infrastructure presents an AAA model for supporting\nsecure global mobility in access networks that are managed by different ad-\nministrators. The notion of a security manager is used through employing\nan AAA broker. The broker acts as a settlement agent, providing security\nand a central point of contact for many service providers (stakeholders).\nThis architecture dynamically provides AAA through forming a virtual layer\non top of the underlying mesh of network domains, thus supporting user\nas well as service mobility across multiple access networks. Through using\nthe DIAMETER protocol [25] in this architecture, the number of security\nassociation required by each mobile node is reduced to only one. Each\nmobile node is just required to have a security association with its home\nAAA server. In addition, by using the roaming capabilities of the DIAME-\nTER protocol, the home DIAMETER server (AAAH) can communicate with\nforeign DIAMETER servers (AAAFs) in other administrative domains. This\narchitecture is illustrated in Figure 8.4. Through the required security asso-\nciation between the AAAH and the mobile node, keys can be created for\neach security association. The keys destined for the foreign and home agent\nare propagated to their nodes via the Diameter protocol, while the key des-\ntined for the mobile node is sent via the MIP protocol [26] resulting in an\nHome\nAAA\nHome network\nCN network\nVisited network\nInternet\nHighest foreign\nagent\nLocal\nAAA\nLocal\nAAA\nHome\nagent\nAAA\nbroker\nAAA\nbroker\nAAA\nbroker\nCorresponding\nnode (CN)\nFigure 8.4\nLightweight AAA infrastructure for mobility support across multiple\ndomains.\n" }, { "page_number": 289, "text": "278\n■\nSecurity in Wireless Mesh Networks\nintegrated MIP/DIAMETER architecture. This AAA infrastructure is useful\nin commercial WMNs deployment allowing dynamic AAA, providing some\nuseful improvements compared to the basic mobility protocol: authentica-\ntion for signaling messages, accounting of network usage, minimal use of\ncryptographic keys, and the non-use of digital signatures.\nThe concept of advanced wireless network architecture is introduced in\n[27] for efficient communications in complex environments, where diffrac-\ntion, attenuation, multi-path, scattering, and fading phenomena are fre-\nquent. A hybrid network architecture using WLAN is proposed that can be\nused for high bandwidth applications such as voice and video snapshots.\nThis architecture is depicted in Figure 8.5. The WLAN APs are connected\nusing a mesh topology while the mobile nodes are to be connected to one\nof the APs using a star topology. The mesh connections between APs allow\nredundant routes that are desirable in dynamic wireless environments. It\nis proposed to use the 802.11f [28] Inter Access Point Protocol (IAPP) to\nhandle mobile nodes hand-offs from one AP to another without losing the\nIP connectivity. Thus, APs need to be connected to a centralized server\nsuch as RADIUS server. The inter-network handoffs is proposed to be han-\ndled using MIP. Applying this architecture in WMNs has two advantages:\n(1) allowing better performance of the AAA process, and (2) providing fast\nsecure roaming. The fact that APs are connected through a mesh topology\nFigure 8.5\nWLAN mesh topology.\n" }, { "page_number": 290, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n279\nfacilitates the exchange of authentication messages between the APs during\nthe authentication or re-authentication of each mobile node. In addition, in\ncase of roaming of a previously authenticated mobile node to a new AP, the\nauthentication process is optimized thanks to the possible communication\nbetween APs. However, the main limitation of this architecture lies in the\nnon-support of multi-hop communication between mobile clients. One way\nto overcome this limitation is by allowing extended mesh topology among\nthe mobile clients. Furthermore, employing the IAPP limits the application\nof this architecture to a specific type of mobile devices. Consequently, an\nalternative solution is needed for general applicability in WMNs; for exam-\nple, broadcasting between APs in ad hoc mode can be a simple means of\ncommunication between APs.\n8.4.4\nExtensible Authentication Protocol Variants\nThe mesh network model with no structure and no trust between the nodes\nmakes the security problem more complex, especially that attackers do not\nneed physical access and they can access layer 2 informations. Also, the\nattacker’s job is easier in terms of finding multiple points of attachments to\nthe network. IEEE 802.1X has been applied to resolve some of the security\nproblems introduced in the 802.11 standard, where the mobile station and\nthe AS authenticate each other through applying an upper layer authenti-\ncation protocol like EAP-TLS (Extensible Authentication Protocol encapsu-\nlating Transport Layer Security) protocol [29] in most of the cases. Although\nEAP-TLS offers mutual authentication, it introduces high latency in WMNs\nbecause each terminal behaves as an authenticator for its neighbor to reach\nthe AS, which can result in longer paths to the AS. Furthermore, in case\nof high mobility of terminals frequent re-authentications due to frequent\nhand-offs can make the network unusable with real-time traffic. Conse-\nquently, variants of EAP are proposed as individual research contributions\nto adapt the 802.1X authentication model to the multi-hop communica-\ntion as well as the WMN environment. This section discusses some recent\nrelated contributions.\n8.4.4.1\nEAP with Token-Based Re-Authentication\nThe dynamic environment together with the multiple possible connectiv-\nities in WMNs raise the need for secure fast hand-off protocols. Because\neach node requiring access to the mesh network initially performs a full\nand costly authentication, then re-using the information of this initial au-\nthentication can speed up the following re-authentications and enhance\nprotocol performance. In this context, a fast secure hand-off protocol is\npresented in [30], which allows mutual authentication and provides access\ncontrol protection through limiting the possibility of insider attackers during\n" }, { "page_number": 291, "text": "280\n■\nSecurity in Wireless Mesh Networks\nthe re-authentication process. To achieve this, old authentication keys are\nremoved from one host to the other. Thus, any host on the network should\nnot receive keys it does not need, but should rather ask for keys from its\nneighbors or from the AS when they are needed.\nThe present solution proposes a token-based re-authentication scheme\nbased on a two-way handshake between the host that performs the hand-\noff and the AS. It is chosen to involve the AS in every hand-off to have a\ncentralized entity for monitoring the network. An authentication token, in\nthe form of keying material, is provided by the authenticator of the network\n(whether an AP or a host in the mesh network) to the AS to obtain the PMK\nkey. Initially, the mobile client performs a full EAP-TLS authentication, gen-\nerating a PMK key that is then shared between the mobile client and its au-\nthenticator. Whenever the mobile client performs a hand-off to another au-\nthenticator, the new authenticator should receive the PMK key to avoid a full\nre-authentication. The new authenticator must issue a request to the AS to\nreceive the PMK, adding to the request a token in the form of cryptographic\nmaterial to prove that it is in contact with the mobile client who owns the\nrequested PMK. Actually, this token is generated by the mobile client while\nperforming the hand-off and is transmitted to the new authenticator. If the\nAS verifies the token, it then issues the PMK to the new authenticator.\nThe fast re-authentication presented in this approach permits central-\nized and hence secure management of the network. However, the need\nto involve the AS with each re-authentication may cause some constraints\nin WMNs in which mobile nodes have random and mostly high dynamic\nbehavior. A distributed-based token verification will be more suitable to\nWMNs, especially for open and multi-hop communication scenarios. Fur-\nthermore, the presented solution does not explain the authentication/\nre-authentication in case of multi-hop communication, which is a liable\nscenario in WMNs. Delegation or distribution of the authenticator’s role\namong mobile clients is a useful solution in such a context.\n8.4.4.2\nEAP-TLS over PANA\nA security architecture suitable for multi-hop mesh network is presented in\n[31], employing EAP-TLS over PANA (Protocol for carrying Authentication\nand Network Access) [32]. This work proposes an authentication solution\nfor wireless mesh networks growing in an ad hoc manner and using ad hoc\nnetwork capabilities. An authentication architecture is developed, and data\nconfidentiality is assured. IEEE 802.1X is adapted so that mobile nodes\ncan be authenticated by mesh access routers that can be APs as well as\nmobile hosts. The authentication between mobile nodes and mesh access\nrouters depending on MAC addresses, according to the 802.1X authentica-\ntion model, requires mobile clients to be directly attached to mesh routers.\nBecause PANA enables clients to authenticate to the access network using\n" }, { "page_number": 292, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n281\nIP protocol, it is used in this work to overcome the problem of associa-\ntion between mobile clients and mesh access routers that can be attached\nthrough more than one intermediate node. Because PANA is an EAP lower\nlayer, any EAP method is suitable for clients’ authentication.\nWhen a new mobile node joins the network, it first gets an IP address\n(pre-PANA address) from a local DHCP server. Then, PANA protocol is\ninitiated so that the mobile node discovers the PANA Access Router (PAA) to\nauthenticate. After successful authentication, the mobile client initiates the\nInternet Key Exchange (IKE) protocol with the mesh router for establishing\na security association. Finally, IPSec tunnel ensures data protection over\nthe radio link and a data access control by the mesh router. During the\nauthentication and authorization phases, PANA uses EAP message exchange\nbetween the client and the PAA, where the PAA relays EAP messages to\nthe AS using EAP over RADIUS. EAP-TLS message is used in this approach;\nhowever, any other application suitable EAP method can be used.\nBecause this solution proposes an architecture which is independent of\nthe wireless media, it is appropriate for heterogeneous WMNs’ future appli-\ncations and in WMNs that are managed by different operators/administrative\ndomains employing similar or different technologies. However, employing\nPANA necessitates the existence of IP addresses among mesh nodes, which\nis still an unsolved problem in the WMN environment.\n8.4.4.3\nEAP-TLS Using Proxy Chaining\nThe contributions of [17] and [33] propose adaptive EAP solutions for au-\nthentication and access control in the multi-hop wireless environment. In\n[17], an adapted EAP-TLS approach is used to allow authentication of mobile\nnodes that do not exist in any AP communication range. A delegation pro-\ncess is used among mobile nodes, through selecting auxiliary authenticators\nin a recursive manner until reaching the AS. To allow extended forward-\ning and exchange of EAP-TLS authentication messages, proxy RADIUS is\ninvolved using proxy chaining among the intermediate nodes between the\nmobile client requesting the authentication and the AS. This approach per-\nmits the storage of mobile clients’ authentication keys among auxiliary au-\nthenticators, which speeds up the re-authentication process and enhances\nthe performance of this adaptive EAP-TLS mechanism. This solution is ap-\nplicable in the WMN environment, especially in scenarios of multi-hop\ncommunication. However, a sort of communication is required between\nauxiliary authenticators to exchange the authentication information con-\ncerning the roaming clients. To support secure roaming across different\nwireless mesh routers (WMRs), communication is required between old and\nnew WMRs during mobile clients’ roaming. This can take place through in-\nstalling central elements/switches linking WMRs and allowing information\ncentralization and distribution between them.\n" }, { "page_number": 293, "text": "282\n■\nSecurity in Wireless Mesh Networks\nAnother adaptive EAP-TLS solution is presented in [33], which is mainly\nproposed for vehicular networks environment; however, it can be useful in\nWMNs. This solution employs a Kerberos authentication server as a central\nserver for all mobile nodes. At a first step, each mobile node should au-\nthenticate to the Kerberos server prior to connection to the network. As a\nresult of this initial authentication, each mobile node obtains a public key\ncertificate for later use in the network. During communication between the\nnodes, each two communicating parties can mutually authenticate using\nEAP-TLS in an ad hoc mode following a client/server model without in-\nvolving the AS, but rather the previously obtained public key certificates\nare used. Employing the Kerberos authentication model in WMNs is use-\nful in managing authorization to different services, especially in case of\nseveral communicating mesh clusters managed by more than one opera-\ntor. WMRs can mutually authenticate through the distributed authentication\napproach proposed; also, this approach is useful for mobile clients authen-\ntication during multi-hop communication that can take place in open WMN\nscenarios. To manage roaming of mobile clients between different WMRs,\ncommunication between WMRs is required. Because mutual authentication\nis possible between WMRs, they can communicate in ad hoc mode to share\nthe authentication information of roaming clients in a secure manner.\n8.4.5\nAAA in Multi-Operator WMNs\nA major objective in WMNs future deployment is services commercializa-\ntion, which will observe a cooperation between different operators and\nservice providers belonging to different administrative domains. However,\nsome challenges need to be resolved to allow ubiquitous services provision\nto mobile clients in such a heterogeneous environment. An important chal-\nlenge concerns the AAA process. Appropriate AAA operation is needed to\npermit wide and scalable WMNs commercial deployment. This necessitates\na trust relationship between operators and providers allowing the contin-\nuous authentication of mobile clients during their roaming across different\nauthentication domains. Roaming of clients between WMRs managed by\ndifferent operators requires authentication of clients each time they con-\nnect to a new operator in a rapid manner with no impact on the continuity\nor the quality of the provided services, especially for real-time applications\nthat are so sensitive to hand-offs delay. Thus trust should exist between the\noperator of the home network to which the clients belongs and the new\noperator which is visited by the mobile client. Trust establishment between\noperators/service providers can take place by signing roaming agreements\nor by using long-term keys shared between the different operators/service\nproviders.\nThe charging and accounting of mobile clients across multiple admin-\nistrative domains should be achieved in a transparent means to services\n" }, { "page_number": 294, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n283\nprovision. Special accounting mechanisms and tailored billing systems\nshould be in place, with appropriate business models considering the\nbenefits of both mobile clients and service providers. In this context, inter-\ndomain accounting is important in assuring service availability and conti-\nnuity. The economic interests require the application of usage-sensitive\nbilling systems based on the gathered accounting information for each\nclient. It is recommended that these systems allow online payment or pre-\npaid tokens. However, processing delay constraints should be considered\nas well as the need for authentication and integrity.\nConsidering WMNs operating in an unlicensed spectrum, another im-\nportant challenge in multi-operators coexistence concerns the spectrum\nsharing. Because the same WMN can be managed by different operators or\nWMNs of different operators can interoperate, the utilization of the same\nunlicensed frequency band by different operators is possible. In such case,\nmobile clients attachment to WMRs is based on the received signal strength\nlevel. Consequently, each operator can authorize its WMRs to transmit using\nthe maximum authorized level to assure that it is heard by the maximum of\nits own mobile clients, which results in a bad WMN performance increasing\nthe interference. Policy agreements should take place between operators\nhandling the spectrum sharing without bad performance effects. Mobile\nclients should freely roam across WMRs of different operators attaching to\nthe one offering the best signal quality irrespective of the operator to which\nthe WMRs belong. This roaming policy is expected to be beneficial for both\noperators and clients. Operators can decrease the transmission power of\ntheir devices while serving an increased set of clients. On the other hand,\nmobile clients can easily discover the closest WMRs and benefit from dif-\nferent services offered by multiple operators.\n8.5\nTrust\nIn commercially deployed WMNs, users do not belong to a common group\nand they do not necessarily trust each other or the different operators. At\nthe same time, each operator does not trust the different users. Because\nWMN deployment is essentially driven by business considerations, trust is\nfundamental in such networks, and any security mechanism requires some\nlevel of trust in its underlying components.\nBuilding and maintaining trust is not an easy task in WMNs. Trust can\nbe defined as the belief of a network element that another network ele-\nment, with which it communicates, is functioning in a way that does not\ndisrupt the network operation/services continuity and according to certain\npredefined rules. However, a trust relation is not symmetric; i.e., if X and\nY are two communicating network elements and X trusts Y, this does not\nimply that Y trusts X, which complicates the problem of trust building. In\n" }, { "page_number": 295, "text": "284\n■\nSecurity in Wireless Mesh Networks\naddition, trust is difficult to quantify or to measure. Consequently, rules\nenforcement by organizations or governmental authorities is sometimes\nnecessary to facilitate trust building between the different communicating\nentities. An example of rules enforcement is the governmental regulation\nof the radio spectrum utilization by network operators. Another example is\nthe control of the mobile devices usage to the radio spectrum by network\noperators. Besides rules enforcement, there is a need for technical mecha-\nnisms deployment to encourage users to some desired behavior during their\nparticipation in the network. These mechanisms can also detect/prevent at-\ntacks caused by nodes misbehavior, and are typically based on security and\ncryptographic techniques.\nThere is a traditional focus on securing routing protocols via ensur-\ning the authenticity of routing messages, aiming to provide transmission\namong trusted elements. However, this approach is insufficient as the key\ncharacteristics of WMNs make it possible for attackers, including malicious\nusers, to add routers, establish links, and advertise routes. In addition, an\nattacker could steal the credentials of a legitimate user or a legitimate user\ncould himself turn malicious, and thereby inject authenticated-but-incorrect\nrouting information into the network. Thus, beyond ensuring the security\nof routing protocols, two important issues worth consideration for trust\nassurance in WMNs environment are:\n1.\nCreating a trust relationship between each pair of communicating\nnodes as well as between nodes on the redundant routing paths\nbetween any communicating parties: Reputation-based mechanisms\ncan help in providing a sort of trust among different network ele-\nments in a distributed manner.\n2.\nSecuring the packet forwarding and dealing directly with the packet\nforwarding misbehavior: A way is needed to securely detect and\nlocalize the source of the packet forwarding misbehavior. Conse-\nquently, the problem of forwarding misbehavior can be solved by\ncontrolling the trouble spot, invalidating the compromised creden-\ntials, or taking offline action through a human interface.\n8.5.1\nUsing Reputation for Building Trust\nBecause future business of WMNs is expected to allow interoperability\namong different operators/service providers, a possible example is the\nintegration of different mesh clusters that belong to different operators/\nservice providers including wireless Internet service providers (WISPs).\nHowever, one of the major problems in this approach is the lack of trust\nbetween the heterogeneous communicating entities that belong to dif-\nferent operators/providers. In this context, reputation-based mechanisms\nseem useful for building up trust between mobile users and the different\n" }, { "page_number": 296, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n285\noperators/providers, and at the same time building trust between mobile\nusers belonging to different administrative domains.\nThe work in [34] treats the problem of interoperability between service\nproviders. A reputation system is developed, using an appropriate trust\nmodel. The trust model considers that the home network of a particular\nservice provider can be the home network for some mobile nodes and a\nforeign network (provider) for other nodes. Thus, the home network of\nany provider could not be considered as an always-trusted element for all\nmobile nodes. Furthermore, a mechanism is presented that can enable ser-\nvice providers to predict the QoS they can offer to mobile nodes according\nto the level of trust.\nThis work is basically developed for WiFi networks; however, it can\nbe adapted to WMNs. Applying this approach in the WMN environment is\nbeneficial in terms of having interoperability between multiple providers in\na secure manner. The reputation-based system can allow mobile nodes to\nevaluate the behavior of service providers and at the same time can allow\nservice providers to authorize mobile users services access according to\ntheir level of trust.\n8.5.2\nDetecting Forwarding Misbehavior\nSecure packet forwarding is an approach to detect malfunctioning among\nthe network elements and estimate a level of trust for each network element\naccording to its forwarding behavior. Although a tool such as traceroute [35]\ncould be used in detecting forwarding misbehavior and identify the offend-\ning mesh routers, an attacker can still treat traceroute packets differently\nor can tamper with the traceroute responses sent by other nodes. A secure\ntraceroute SecTrace protocol [36] is developed to securely trace the existing\ntraffic paths. SecTrace allows intermediate routers to prove the traffic recep-\ntion rather than using implicit responses. In addition, SecTrace responses\nare authenticated to verify their origin and prevent spoofing and tampering.\nSecTrace is recommended for the community WMN environment to moni-\ntor end-to-end connectivity to other mesh nodes and to detect connectivity\nproblems.\nThe operation of SecTrace, as in normal traceroute, takes place in a hop-\nby-hop manner to identify the offending routers. Each node on the path is\nbeing asked to respond to traceroute traffic, where each responding node\nprovides a next-hop router identity for the packet in addition to its own\nidentity. A shared key is established by the tracing node prior to sending\nthe traceroute packets, where this key is used to encrypt and authenticate\nthe communication to and from the expected next node. In replying to a\nSecTrace packet, a node sends some agreed-upon identifying marker for\nthe packet to prove to the tracing node that the packet has been received.\nAlso, a strongly secure Message Authentication Code (MAC) is contained in\n" }, { "page_number": 297, "text": "286\n■\nSecurity in Wireless Mesh Networks\nthe reply packet, ensuring its authentic origin. After replying to SecTrace,\nthe replying node becomes the next node for the next step of traceroute.\nSecTrace is useful in the context of deployable WMNs to detect and lo-\ncalize the cause of packet forwarding misbehavior, because securing rout-\ning only is insufficient in such environment. An implementation of SecTrace\n[36] in a WMN scenario shows that it has a negligible performance over-\nhead, making it suitable for monitoring of end-to-end paths and estimating\na trust level for each contributing network element, whether it is a mobile\nclient or a mesh router.\n8.5.3\nTrusted Routing\nMesh networks rely on participation and cooperation of nodes within the\nnetwork during the routing process. However, the fact that participating\nnodes are controlled by different owners, nodes may choose to act in their\nown interest in a way that can impact the networking functioning. In this\ncontext, trusted routing is beneficial in providing additional security in this\nopen environment by allowing each mesh node to prove its identity and\nintegrity.\nThe work in [37] presents a contribution to trusted routing, which ex-\ntends the Ad hoc On-demand Distance Vector (AODV) [38] routing protocol\nto ensure that only trustworthy nodes participate in the network. A system\nis presented that uses trusted computing to prevent selfish or malicious\nnodes from participating in the network. A new protocol named Trusted\nComputing Ad hoc On-demand Distance Vector (TCAODV) has been de-\nveloped to enhance AODV protocol through preventing network abuse by\nselfish and malicious nodes. In TCAODV, a public key certificate is used\nby each node, which is stored within a trusted root used for the purposes\nof routing. The node broadcasts this certificate with Hello messages, where\nneighbors receiving this certificate first verify it through the signature of the\nissuer, then store it as the broadcaster’s public key in case of validation.\nThe RREQ packet sent by each node is signed with a sealed signature,\nusing integrity metrics from the routing module of the sender. The node\nthat receives the RREQ verifies the signature through using the previously\nreceived key for the requester node, and determines if the provided mea-\nsurements are trustworthy. When the destination is not directly reachable\nby the RREQ, the intermediate node strips off the signature, replacing it\nby its own signature and integrity measurements. In addition, a per-route\nsymmetric encryption key is established to ensure that only trusted nodes\nalong the path can use the route. All traffic sent along the route is encrypted\nusing this symmetric key. The TCAODV approach has less overhead on the\nnetwork and can be applied in WMN scenarios. A typical scenario example\nis a community wireless mesh network among houses in residential areas.\nIn this scenario, houses are equipped with wireless nodes that forward\n" }, { "page_number": 298, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n287\ntraffic toward a wired Internet connection, and in turn may also make use\nof this connection.\n8.6\nPrivacy\nPrivacy provision is an important issue worth consideration to widen WMN\ndeployment. Privacy concerns hiding the transferred messages/critical data\nfrom unauthorized parties, which is an important means for controlling\nmessage transfer in WMN environments. However, privacy is difficult to\nachieve even if messages are protected, as there are no security solutions\nor mechanisms which can guarantee that data is not revealed by the autho-\nrized parties themselves. Thus, complementary solutions are important to\nbe in place. Also, communication privacy could not be assured with mes-\nsages protection, as attackers could still observe who is communicating\nwith whom as well as the frequency and duration of the communication\nsessions. This makes personal information susceptible to disclosure. Fur-\nthermore, mobile clients in WMNs can be easily monitored/traced in terms\nof their presence, which causes the exposure of their personal life. Unau-\nthorized parties can learn the mobile clients’ positions/locations through\nobserving their communication. Consequently, there is a need to ensure\nlocation privacy in WMNs.\nTo control the usage of personal information and the disclosure of per-\nsonal data, different types of information hiding from unauthorized parties\nappear to be efficient. The following approaches can be useful in informa-\ntion hiding, depending on what is needed to be protected:\n■\nAnonymity: This is concerned with hiding the identity of the mes-\nsage sender or the message receiver or both of them. In fact, hiding\nthe identity of both the sender and the receiver of the message can\nassure communication privacy. Thus, attackers observing transmis-\nsions could not know who is communicating with whom, thus no\npersonal information is disclosed.\n■\nConfidentiality: This is concerned with hiding the transferred mes-\nsages themselves. Instead of hiding the identity of the sender and\nthe receiver of a message, the message itself is hidden.\n■\nUsing pseudonyms: This is concerned with replacing the identity of\nthe sender and the receiver of the message by pseudonyms which\nfunction as identifiers. Thus, pseudonyms can be used as a reference\nto the communicating parties without hurting their privacy, which\nhelps to assure untraceability of clients. However, it is important to\nassure the unlinkability of pseudonyms and real identifiers.\nThis section discusses privacy protection in WMNs, highlighting some\ninteresting research contributions.\n" }, { "page_number": 299, "text": "288\n■\nSecurity in Wireless Mesh Networks\n8.6.1\nEfficient Key Distribution for Message Protection\nPower-efficient encryption and decryption can achieve message protection\nin WMNs. There is a need for simple, robust, and lightweight security mech-\nanisms that are suitable to the WMN environment and nodes characteristics.\nAlthough the second part of the IEEE 802.11i standard uses Advanced En-\ncryption Standard (AES) protocol to overcome the significant processing\non every packet caused by the previously used Temporal Key Integrity\nProtocol (TKIP), the AES also adds an overhead of eight octets on every\npacket and can still be very expensive. In this context, the contribution of\n[39] presents a State-Based Key Hop (SBKH) protocol that provides a strong\nand lightweight encryption scheme suitable for battery operated devices. It\nis shown that integrating SBKH with 802.11 allows a power and processing\ncost that is much lower than 802.11i encryption mechanisms. SBKH is based\non the concept of state-based encryption, where it does not reinitialize RC4\nstate for every packet. Instead, the same RC4 seed is maintained for a du-\nration that is known to the communicating nodes. The initialization of the\nRC4 state is only carried out when the base key changes. SBKH allows mo-\nbile nodes to be state synchronized, where they keep using the same cipher\nstream to encrypt and decrypt packets exchange between them. In fact, ap-\nplying this scheme in WMNs is important in terms of providing cheap and\nrobust security without additional encryption overhead together with sav-\ning significant processing power, especially for applications of large packet\nsizes. Furthermore, operating with the existing hardware as well as the\nexisting 802.11 protocols is important to millions of 802.11 cards shipped,\nwhere a change in the hardware will not solve the security issues with these\nexisting 802.11 cards.\nThe messages generated in WMNs are sent using multi-hop communica-\ntion among WMRs and mobile clients relaying the messages. Consequently,\nthe use of public key cryptography is a heavy process introducing important\ndelays, and thus leading to sub-optimal utilization of network resources.\nA possible solution consists in establishing or pre-defining secret keys be-\ntween mesh routers that can be used in encrypting messages transferred\nthrough the hop-by-hop communication. However, a major problem in\nWMNs is the distribution of secret keys. To meet the constraints of high and\nunpredictable mobility together with limited power and storage resources\nof mobile nodes, particular key distribution protocols are needed taking\ninto account these constraints and maintaining a strong security level. A\nnew approach for random key pre-distribution is proposed in [40], achiev-\ning both efficiency and security objectives. This work replaces the use of a\nkey pool for random keys by a developed key-generation technique. In this\ndeveloped technique, a large number of random keys can be represented\nby a small number of key-generation keys. Consequently, instead of storing\n" }, { "page_number": 300, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n289\na large number of random keys, each mobile node stores a small number\nof key-generation keys while computing the shared secret keys during the\nbootstrapping phase. This solution is useful in WMN scenarios because it is\nscalable to large network sizes. The distributed solution for secret sharing is\nappropriate for WMN multi-hop communications, whether through WMR\nrelays or mobile client relays. Furthermore, applying this scheme in the\nWMN environment allows a significant reduction in storage requirements,\nwhile maintaining the required security strength.\n8.6.2\nTraffic Privacy\nTraffic preservation is a useful approach in providing communication pri-\nvacy. Despite the necessity of traffic preservation, limited research has been\nconducted on this issue. Indeed, in a community mesh network, the traf-\nfic of mobile users can be observed by the mesh routers residing at its\nneighbors, which could reveal sensitive personal information. A mesh net-\nwork privacy-preserving architecture is presented in [41]. This work targets\ntraffic confidentiality, aiming at deducing the traffic information, such as\nwho the user is communicating with, and the amount and time of traffic. A\nlightweight traffic privacy-preserving mechanism for WMNs is developed,\nbased on the concept of traffic pattern concealment via routing control,\nusing the intrinsic WMN redundancy in terms of multi-paths. As illustrated\nin Figure 8.6, the traffic from the source (gateway) to the destination (mesh\nrouter) is split to many paths, thus all the relaying nodes along the paths\ncould only observe a portion of the entire traffic. Furthermore, the traffic\ncan be split in a random way (spatially and temporally) so that an interme-\ndiate node can have little knowledge to figure out the overall traffic pattern,\nallowing the traffic pattern to be concealed.\nThe present work first defines an information-theoretic metric, then pro-\nposes a penalty-based routing algorithm to allow traffic pattern hiding by\nexploiting the multiple available paths between any two mesh nodes. The\nsource routing scheme is adopted which allows a node to easily learn the\ntopology of the WMN that it belongs to through each received packet,\nwhile the source and destination ID are encrypted. This work can assure\ncommunication privacy in WMNs, where each destination is able to consis-\ntently limit the proportion of mutual information it shares with the observing\nnode. This approach needs more adaptation for the WMN environment and\napplications. The fact of splitting traffic on multiple paths may impact the\ntransmission delay. This can be harmful to the continuity of service of real-\ntime applications, such as VoIP and streaming, which are delay sensitive.\nFurthermore, when applying this approach in WMN scenarios with multi-\nhop communication among mobile clients, multi-path transmission among\n" }, { "page_number": 301, "text": "290\n■\nSecurity in Wireless Mesh Networks\nInternet\nMulti-paths for data delivery\nTo the mesh router\nMesh router\nGateway\nFigure 8.6\nPreserving traffic privacy.\nmobile nodes (relays) can cause packets loss, which in turn impacts the\ntransmission quality. Consequently, positioning information of the relaying\nmobile clients is important to be acquired to select the relaying multi-paths\naccording to their mobility behavior and patterns.\n8.6.3\nNon-Traceability\nIn fact, the behavior of mesh nodes can be easily traced by adversaries due\nto the use of wireless channels, multi-hop connections through intermedi-\nate nodes, and convergence of traffic to WMRs. Hiding nodes activity is an\napproach that can prevent nodes traceability, assuring their privacy. Cryp-\ntographic approaches are not appropriate to achieve nodes privacy in terms\nof hiding nodes activities, as they are not efficient in case of internal attack-\ners among the WMRs or the mobile clients. At the same time, redundancy\nin transmissions through broadcasting at WMRs or gateways can hide the\nactivity of the receiver node; however, an internal attacker can discover the\nnode when it sends a message to a WMR or a gateway. In [42], a solution\nis proposed with the objective of hiding an active node that connects to a\n" }, { "page_number": 302, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n291\ngateway router, where this active mesh node has to be anonymous. A novel\ncommunication protocol is designed to protect nodes privacy using both\ncryptography and redundancy. This protocol uses the concept of onion\nrouting in wired networks [43], adapting it to the WMN environment. In\nthis solution, an end user requiring an anonymous communication sends\na request to an onion router (OR). The OR acts as a proxy for the mo-\nbile user, and the communication between the end user/mobile client and\nthe OR is protected from adversaries. The proxy constructs a route con-\nsisting of other ORs and constructs an onion using the public keys of the\nrouters on the route. The onion is constructed such that the most inner part\nis the message for the intended destination, and the message is wrapped\nby being encrypted using the public keys of the ORs in the route with\ntheir same order in the route. The ID of the session initiator is not carried\nin the constructed route, where the initiator is kept anonymous to other\nmesh nodes. To prevent attackers from monitoring routes from gateways\nto initiator nodes, the constructed route between the initiator node and the\ngateway does not end at the initiator; however, it extends for a few extra\nhops carrying dummy information generated by the initiator node.\nThis work protects the routing information from insider and outsider\nattackers, making each node behavior/activity undistinguishable. However,\nthere should be a trade-off between the anonymity and the computing/\ncommunication overhead. It should be assured that achieving a higher level\nof anonymity should not result in higher overhead cost.\n8.7\nConclusion and Outlook\nTo further ensure security of WMNs, some essential strategies need to be\nconsidered. Security and privacy mechanisms and architectures for access\nnetworks including WMNs have considered the lower layers in the form of\nsecurity over wireless networks and the upper service layers in the form\nof application and transport security. However, what is still missing is a\ngeneral solution which is both adaptable to the network types and also\ntakes into account end-system capabilities as well as enabling inter-domain\nAAA negotiation.\nSecurity mechanisms need to be embedded into MAC protocols to detect\nand prevent misbehavior in channel access and into network protocols\nproviding a secure routing. Moreover, new or adaptive upper layer proto-\ncols are needed for WMNs, taking into consideration centralized and opened\nWMN scenarios together with the multi-hop communication principle.\nGenerally, multi-layer security is desired as attacks occur simultaneously\nin different protocol layers. It might be important to develop cross-layer\n" }, { "page_number": 303, "text": "292\n■\nSecurity in Wireless Mesh Networks\nframework for security monitoring to detect attacks responding quickly to\nthem. Furthermore, it is necessary to provide sufficient authentication for\nuser nodes to authenticate mesh nodes or for a downstream mesh node to\nauthenticate an upstream mesh node. However, it is important to be mind-\nful of the overhead caused by authentication as wireless users or mesh\nnodes are often constrained by limited battery power, computing power,\nor memory space. Also, unacceptable authentication delay might impact\nservice continuity. The future deployment of WMNs will observe multi-\noperators’ coexistence, which requires appropriate AAA systems that allow\nmobile clients authentication and accounting across multiple administrative\ndomains.\nProviding trust between mesh nodes is an important aspect; however,\nthe domain of WMNs still lacks appropriate mechanisms capable of intro-\nducing trusted elements. In this context, upper layer architectures for trust\nprovision between nodes should be provided, taking into account the spe-\ncial characteristics of WMNs. Consequently, an important issue that should\nalso be considered is the measurement and estimation of the trust levels\nbetween nodes. Appropriate metrics should be developed for calculating\ntrust levels in WMNs at lower layers. As well, new architectures for trust\ninfrastructure assurance are needed at the application level.\nThe open medium property of WMNs makes them vulnerable to privacy\nattacks. The behavior of mesh nodes can be easily monitored and traced by\nadversaries due to the use of wireless channel, multi-hop communication,\nand traffic convergence to mesh routers. Despite the necessity of privacy to\nprotect sensitive personal information and prevent client traceability, lim-\nited research contributions have been conducted toward privacy preserving\nin WMNs. This subject still needs wide investigation and studies, and could\nimpact the type of applications in future WMN deployment.\nFor future deployment of WMNs, further important open issues are still\nnot covered and need more investigation from the research community as\nwell as the industry. One important issue is the secure auto-configuration\nof mobile nodes in this environment. Another issue is the fast and secure\nassociation between mobile nodes in a totally distributed manner and with\nhigh mobility that is mostly taking place in WMN open scenarios. In provid-\ning intelligent commercial WMN services, an interesting point to be studied\nis employing rewarding mechanisms, in terms of providing incentives to\nmobile nodes to cooperate, as a means of accounting mobile users. Finally,\napplying the Grid Computing paradigm seems useful in WMNs, in terms of\naggregating the mesh nodes resources to carry out heavy security services.\nA wide take up in Grid Computing is the appropriate security models and\nthe cross-organizational AAA for collaborative business. In this new trend,\nmobile users with varying context and capabilities act as resource providers\nand at the same time clients participating in the grid.\n" }, { "page_number": 304, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n293\nReferences\n[1]\nM. Kodialam and T. Nandagopal, Characterizing the Capacity Region in\nMulti-Radio Multi-Channel Wireless Mesh Networks, MobiCom, 2004.\n[2]\nH. Moustafa, U. Javaid, T. M. Rasheed, S. M. Senouci, and D. Meddour,\nA Panorama on Wireless Mesh Networks: Architectures, Applications and\nTechnical Challenges, Wimeshnets 2006, 2006.\n[3]\nP. Krishnamurthy, D. Tipper, and Y. Qian, The Interaction of Security and\nSurvivability in Hybrid Wireless Networks, IEEE International Conference\non Performance, Computing, and Communications, 2004.\n[4]\nS. Marti, T. J. Giuli, K. Lai, and M. Baker, Mitigating Routing Misbehavior in\nMobile Ad Hoc Networks, MobiCom, 2000.\n[5]\nS. Buchegger and J.-Y. Le Boudec, Performance Analysis of the CONFI-\nDANT Protocol (Cooperation Of Nodes: Fairness In Dynamic Ad-hoc NeT-\nworks), MobiHoc, 2002.\n[6]\nR. Mahajan, M. Rodrig, D. Wetherall, and J. Zahorjan, Sustaining Cooper-\nation in Multi-Hop Wireless Networks, Second Symposium on Networked\nSystems Design and Implementation (NSDI ’05), 2005.\n[7]\nP. Kyasanur and N. H. Vaidya, Detection and Handling of MAC Layer Mis-\nbehavior in Wireless Networks, International Conference on Dependable\nSystems and Networks (DSN ’03), 2003.\n[8]\nM. Raya, J. P. Hubaux, and I. Aad, Domino: A System to Detect Greedy\nBehavior in IEEE 802.11 Hotspots, Second International Conference on\nMobile Systems, Applications and Services (MobiSys 2004), 2004.\n[9]\nS. M. Faccin, C. Wijting, J. Kneckt, and A. Damle, Mesh WLAN networks:\nConcept and system design, IEEE Wireless Communication, April 2006.\n[10]\nH. Deng, W. Li, and D. Agrawal, Routing security in wireless ad hoc net-\nworks, IEEE Communication Magazine, October 2002.\n[11]\nM. Barbeau, WiMax/802.16 Threat Analysis, First ACM International Work-\nshop on Quality of Service and Security in Wireless and Mobile Networks\n(Q2SWinet ’05), 2005.\n[12]\nI. Akylidiz, X. Wang, and W. Wang, Wireless mesh networks: A survey,\nComputer Networks—Elsevier Science, No. 47, January 2005.\n[13]\nA. Sikora, Design challenges for short-range wireless networks, IEEE WLAN\nSystems and Internetworking, Vol. 151, No. 5, October 2004.\n[14]\nIEEE Std 802.11i, Medium Access Control Security Enhancements, 2004.\n[15]\nIEEE Std 802.1X, Local and Metropolitan Area Networks Port-Based Net-\nwork Access Control, 2001.\n[16]\nX. Zheng, C. Chen, C.-T. Huang, M. Matthews, and N. Santhapuri, A Dual\nAuthentication Protocol for IEEE 802.11 Wireless LANs, IEEE Second Inter-\nnational Symposium on Wireless Communication Systems, 2005.\n[17]\nH. Moustafa, G. Bourdon, and Y. Gourhant, Authentication, Authorization\nand Accounting (AAA) in Hybrid Ad hoc Hotspots Environments, Fourth\nACM International Workshop on Wireless Mobile Applications and Services\non WLAN Hotspots (WMASH ’06), 2006.\n[18]\nC. Rigney, S. Willens, A. Rubins, and W. Simpson, Remote Authentication\nDial In User Service (RADIUS), RFC 2865, June 2000.\n" }, { "page_number": 305, "text": "294\n■\nSecurity in Wireless Mesh Networks\n[19]\nB. Aboba and J. Vollbrecht, Proxy Chaining and Policy Implementation in\nRoaming, RFC 2607, June 1999.\n[20]\nS. Zhu, S. Xu, S. Setia, and S. Jajodia, LHAP: A Lightweight Hop-by-\nHop Authentication Protocol for Ad-hoc Networks, IEEE 23rd International\nConference on Distributed Computing Systems Workshops (ICDCSW ’03),\n2003.\n[21]\nS. Zhu, S. Xu, S. Setia, and S. Jajodia, LHAP: A lightweight network access\ncontrol protocol for ad hoc networks, Elsevier Ad Hoc Networks Journal,\nVol. 4, No. 5, September 2006.\n[22]\nL. Lamport, Password Authentication with Insecure Communication, Com-\nmunications of the ACM, Vol. 24, No. 11, November 1981.\n[23]\nA. Perrig, R. Canetti, D. Song, and J. Tygar, Efficient and Secure Source Au-\nthentication for Multicast, Network and Distributed Security System Sym-\nposium (NDSS ’01), 2001.\n[24]\nN. Prasad, M. Alam, and M. Ruggieri, Light-weight AAA infrastructure for\nmobility support across heterogeneous networks, Wireless Personal Com-\nmunications, Vol. 29, 2004.\n[25]\nDIAMETER: http://www.diameter.org/\n[26]\nC. Perkins, Mobile networking through Mobile IP, IEEE Internet Computing,\nVol. 2, Issue 1, January/February 1998.\n[27]\nK. Srinivason, M. Ndoh, and K. Kaluri, Advanced Wireless Networks for Un-\nderground Mine Communications, First International Workshop on Wireless\nCommunications in Underground and Confined Areas (IWWCUCA 2005),\n2005.\n[28]\nIEEE Std 802.11f, IEEE Trial-use Recommended Practice for Multi-vendor\nAccess Point Interoperability via an Inter-access Point Protocol Across Dis-\ntribution Systems Supporting 802.11 Operation, 2003.\n[29]\nB. Aboba and D. Simon, PPP EAP TLS Authentication Protocol, RFC 2716,\n1999.\n[30]\nR. Fantacci, L. Maccari, T. Pecorella, and F. Frosali, A Secure and Performant\nToken-based Authentication for Infrastructure and Mesh 802.1X Networks,\nInfoCom, 2006.\n[31]\nO. Cheikhrouhou, M. Maknavicius, and H. Chaouchi, Security Architecture\nin a Multi-hop Mesh Network, Fifth Conference on Security and Network\nArchitectures (SAR), 2006.\n[32]\nM. Parthasarathy, Protocol for Carrying Authentication and Network Access\n(PANA) Threat Analysis and Security Requirements, RFC 4016, March 2005.\n[33]\nH. Moustafa, G. Bourdon, and Y. Gourhant, Providing Authentication and\nAccess Control in Vehicular Network Environment, 21st IFIP TC-11 Inter-\nnational Information Security Conference (IFIP/SEC 2006), 2006.\n[34]\nN. B. Salem, J-P. Hubaux, and M. Jakobsson, Reputation-based Wi-Fi de-\nployment, Mobile Computing and Communication Review, Vol. 9, No. 3,\n2005.\n[35]\nV. Jacobson, The Traceroute Manual Page, Lawrence Berkeley Laboratory,\nBerkeley, CA, December 1988.\n" }, { "page_number": 306, "text": "Providing Authentication, Trust, and Privacy in Wireless Mesh Networks\n■\n295\n[36]\nG. Mathur, V. Padmanabhan, and D. Simon, Securing Routing in Open Net-\nworks Using Secure Traceroute, Microsoft Research technical report (MSR-\nTR-2004-66), July 2004.\n[37]\nM. Jarrett and P. Ward, Trusted Computing for Protecting Ad-hoc Routing,\nIEEE Fourth Annual Communication Networks and Services Research Con-\nference (CNSR 2006), 2006.\n[38]\nC. Perkins, E. Belding-Royer, and S. Das, Ad hoc On-demand Distance\nVector (AODV) Routing, RFC 3561, July 2003.\n[39]\nS. Michell and K. Srinivasan, State Based Key Hop Protocol: A Lightweight\nSecurity Protocol for Wireless Networks, ACM International Workshop on\nPerformance Evaluation of Wireless Ad hoc, Sensor and Ubiquitous Net-\nworks (PE-WASUN ’04), 2004.\n[40]\nK. Ren, K. Zeng, and W. Lou, A new approach for random key pre-\ndistribution in large-scale wireless sensor networks, Wireless Communi-\ncations and Mobile Computing, Vol. 6, 2006.\n[41]\nT. Wu, Y. Xue, and Y. Cui, Preserving Traffic Privacy in Wireless Mesh\nNetworks, 2006 International Symposium on a World of Wireless, Mobile\nand Multimedia Networks (WoWMoM 2006), 2006.\n[42]\nX. Wu and N. Li, Achieving Privacy in Mesh Networks, The Fourth ACM\nWorkshop on Security of Ad Hoc and Sensor Networks (SASN ’06), 2006.\n[43]\nM. Reed, P. Syverson, and D. Goldschlag, Anonymous connections and\nonion routing, IEEE Journal on Selected Areas in Communication, special\nissue on copyright and privacy protection, 1998.\n" }, { "page_number": 307, "text": "" }, { "page_number": 308, "text": "Chapter 9\nNon-Interactive Key\nEstablishment in Wireless\nMesh Networks1\nZhenjiang Li and J.J. Garcia-Luna-Aceves\nContents\n9.1\nIntroduction ......................................................... 298\n9.2\nBasics of the Self-Certified Key Cryptosystem ..................... 301\n9.3\nNon-Interactive Key Agreement and Progression .................. 303\n9.3.1\nS-NIKAP and A-NIKAP ...................................... 303\n9.3.2\nApplication Scenarios of NIKAP ............................ 305\n9.4\nAd hoc On-demand Secure Routing Protocol ...................... 306\n9.4.1\nAssumptions ................................................. 306\n9.4.2\nRoute Discovery ............................................. 306\n9.4.2.1\nRoute Request Initialization ........................ 306\n9.4.2.2\nRoute Request Forwarding ......................... 308\n9.4.2.3\nChecking RREQ at Destination D .................. 308\n9.4.3\nRoute Maintenance .......................................... 308\n9.5\nSecurity Analysis..................................................... 310\n9.6\nPerformance Evaluation ............................................. 313\n1 This work was supported in part by the Baskin Chair of Computer Engineering at\nUCSC, the National Science Foundation under Grant CNS-0435522, and the U.S. Army\nResearch Office under Grant no. W911NF-05-1-0246. Any opinions, findings, and con-\nclusions are those of the authors and do not necessarily reflect the views of the funding\nagencies.\n297\n" }, { "page_number": 309, "text": "298\n■\nSecurity in Wireless Mesh Networks\n9.7\nRelated Work and Open Issues ..................................... 317\n9.8\nConclusion........................................................... 319\nReferences................................................................. 320\nSymmetric cryptographic primitives are preferable in designing security pro-\ntocols for wireless mesh networks (WMNs) because they are computation-\nally affordable for resource-constrained mobile devices forming a WMN.\nMost proposed key-establishment schemes for symmetric cryptosystems\nassume services from a centralized authority (either online or offline), or\ninvolve interaction between communicating parties. However, requiring\naccess to a centralized authority, or ensuring that correct routing be estab-\nlished before the key agreement is done, is difficult to attain in wireless\nnetworks.\nWe present a new non-interactive key agreement and progression\n(NIKAP) scheme for wireless networks, which does not require an on-\nline centralized authority, can establish and update pairwise shared keys\nbetween any two nodes in a non-interactive manner, is configurable to op-\nerate synchronously (S-NIKAP) or asynchronously (A-NIKAP), and has the\nability to provide differentiated security services wireless routers the given\nsecurity policies. As the name implies, NIKAP is especially valuable to sce-\nnarios in which shared secret keys are desired to be computed without\nnegotiation between mobile nodes over insecure channels, and also need\nto be updated frequently.\nAs an application example, we present the Ad hoc On-demand Secure\nRouting (AOSR) protocol based on NIKAP to secure the signaling of on-\ndemand ad hoc routing, which exploits pairwise keys between pairs of\nnodes and hash values keyed with them to verify the validity of the path\ndiscovered. Analysis and simulation results show that AOSR has low com-\nmunication overhead caused by the key establishment process due to the\nuse of NIKAP, effectively detects or thwarts a wide range of attacks to on-\ndemand ad hoc routing, and is able to maintain a high packet-delivery ratio,\neven when a considerable percentage of nodes are compromised.\n9.1\nIntroduction\nA wireless mesh network (WMN) is a dynamically self-organized network of\nwireless nodes that automatically establish and maintain mesh connectivity\namong themselves (forming, in effect, an ad hoc network). A WMN consists\nof mesh routers and mesh clients, and each node operates not only as a\nhost, but also as a router that forwards packets for other nodes. This feature\nenables advantages such as low operation cost, robustness, and extendable\n" }, { "page_number": 310, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n299\nservice coverage. However, the ad hoc deployment without centralized\nadministration and the highly dynamic nature of wireless networks also\nbring up new challenges to systems built on them, among which security\nis a pressing problem.\nIn general, there are three cryptographic techniques that can be used to\ndevise security mechanisms for WMNs: one-way hash functions, symmetric\ncryptosystems, and asymmetric (or public key) cryptosystems. An asym-\nmetric cryptosystem is more efficient in key utilization in that the public\nkey of a node can be used by all the other nodes; a symmetric cryptosystem\nrequires the existence of a shared key between two communicating nodes.\nHash functions can be implemented quickly, and usually work together\nwith symmetric or asymmetric algorithms to create more useful credentials,\nsuch as a digital certificate or a keyed hash value (i.e., a keyed message\nauthentication code).\nPortable devices forming a WMN usually have limited battery life and\nmust share a relatively limited transmission bandwidth. Therefore, sym-\nmetric cryptosystems are preferable in ad hoc scenarios due to their com-\nputational efficiency (conducting an asymmetric algorithm usually is three\nor four orders of magnitude slower than the symmetric counterpart). For\na symmetric cryptosystem to work, a shared key must be established be-\ntween each pair of communicating entities. The key establishment problem\nbetween two network principals is well understood for conventional com-\nmunication networks, and generally can be resolved by key distribution or\nkey agreement.\nThe classic key-distribution scheme, such as Kerberos [1], requires an\nonline centralized authority (CA) to generate and distribute keys for nodes.\nHowever, this is not suitable for WMNs. In practice, the online CA can be\nunavailable to some of the nodes, or even the whole network during cer-\ntain time periods, because of the unpredictable state of wireless links and\nnode mobility. Given that the CA is the single point of failure, compromis-\ning the CA jeopardizes the security of the entire system. More importantly,\nthe Kerberos system is designed to provide authentication and key distri-\nbution services for networks structured based on the client/server model,\nwhich, however, is not the case of WMNs. In WMNs, nodes are assumed\nto be willing to route packets for other nodes and behave as peers of one\nanother, such that every node has the responsibility of a mobile router in\naddition to a common network user. Therefore, a WMN is a peer-to-peer\ncommunication system for the purpose of routing, into which the conven-\ntional client/server model-oriented, centralized key distribution approach\ndoes not fit. Recently proposed key distribution protocols [2] for wireless\nenvironments replace the functionality of CA by a subset of nodes in the\nnetwork. However, this approach still relies on a small number of nodes,\nand it is not clear whether sharing the CA functionality among multiple\nnodes can perform better than using a single CA, given that applications\n" }, { "page_number": 311, "text": "300\n■\nSecurity in Wireless Mesh Networks\nneed to contact multiple nodes that can be multiple hops away, to obtain\nthe desired keys.\nKey agreement protocols, such as the Diffie–Hellman key exchange pro-\ntocol [3] and many variations derived from it, do not need an online CA\nand compute the shared keys between nodes on-demand. These protocols\nare interactive schemes in that nodes need to exchange messages between\nthem to establish the desired keys, for which active routes must pre-exist\nfor such approaches to work. The assumption of pre-existing routes be-\ntween two communicating parties, which may be multiple hops away from\neach other, contradicts the need to secure the routing discovery process\nbetween such nodes in the first place. Even if such an assumption is sat-\nisfied, network dynamics can tear routes down in the middle of the key\nnegotiation, and as such no key can be agreed upon. Moreover, interactive\nkey agreement protocols are not scalable in terms of communication over-\nhead, because messages exchanged for key establishment can consume\nsignificant CPU cycles and wireless bandwidth in such a highly dynamic\nenvironment as WMNs, which can become even worse if the shared keys\nbetween nodes need to be updated frequently.\nMotivated by the observations above and based on self-certified key\n(SCK) [4] cryptosystem, we propose new NIKAP protocols to facilitate the\nkey agreement process in WMNs. In NIKAP-oriented protocols, pairwise\nkeys can be computed between two nodes in a non-interactive manner, as\nwell as the subsequent key progression (rekeying) process. NIKAP needs\nthe aid of a CA only at the initial network formation, and the CA can be\nentirely offline thereafter. Consequently, single-point failures are avoided\nduring the operation of the deployed WMN. Compared with other key dis-\ntribution and agreement approaches, NIKAP saves scarce energy and band-\nwidth of wireless nodes in transmitting, receiving, and processing messages.\nTo our knowledge, NIKAP is the first key establishment scheme that sup-\nports the non-interactive key agreement and subsequent key progression\nsimultaneously. Though there are a few protocols that can establish shared\nkeys between nodes non-interactively based on either matrix threshold\nkey pre-distribution (MTKP), or polynomial threshold key predistribution\n(PTKP) [16], none of them supports non-interactive key progression.\nThe rest of the chapter is organized as follows. For completeness, Sec-\ntion 9.2 reviews the basic idea of the SCK cryptosystem, which was first\nintroduced by Petersen and Horster [4]. Section 9.3 presents S-NIKAP and\nA-NIKAP, the non-interactive key agreement and progression protocols tai-\nlored for WMNs, in which we also discuss scenarios to which NIKAP-based\nprotocols can be applied. Section 9.4, Section 9.5, and Section 9.6 present\nthe results of our recent use of NIKAP to secure the routing process in wire-\nless ad hoc networks. We compare NIKAP with other key distribution and\nagreement approaches proposed for wireless environments in Section 9.7,\nand present the concluding remark in Section 9.8.\n" }, { "page_number": 312, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n301\n9.2\nBasics of the Self-Certified Key Cryptosystem\nIn an asymmetric cryptosystem, there are two ways of ensuring the au-\nthenticity of a public key: explicit verification and implicit verification. In\nexplicit verification, a trusted centralized authority signs a certificate that\nbinds a public key and the identity (ID) of its owner. Then any user can\nverify the certificate explicitly provided that the public key of the central-\nized authority is known. In implicit verification, the authenticity of a public\nkey is verified when it is used for encryption (or decryption), signature ver-\nification, key exchanging, or other cryptographic operations. For example,\na successful verification of a signature means that the public key matches\nthe private key used to construct this signature. A self-certified key (SCK)\nsystem follows the track of implicit verification. In the following, we first\nsummarize the basic primitives used by SCK to establish and update the\nshared pairwise keys between two communicating parties. In such cases,\nthe authenticity of a public key is verified when the shared keys derived\nbased on it are used, for example, to encrypt and decrypt data, and to\ngenerate and check keyed hash values.\n■\nInitialization: A CA Z is assumed to exist before the network forma-\ntion. Z chooses large primes p, q with q|(p −1) (i.e., q is a prime\nfactor of p−1), a random number kA ∈Z∗\nq, where Z∗\nq is a multiplica-\ntive sub-group with order q, and generator α; then Z generates its\n(public, private) key pair (xZ, yZ). We assume that the public key yZ\nis known to every node that participates in the network. To issue the\nprivate key for node A with identifier IDA, Z computes the signature\nparameter rA = αkA (mod p) and sA = xZ · h(IDA, rA) + kA (mod q),\nwhere h(·) is a collision-free one-way hash function and (mod p)\nmeans modulo p. Node A publishes the parameter rA, called the\nguarantee, together with its identifier IDA, and keeps xA = sA as its\nprivate key. The public key of A can be computed by any node that\nhas yZ, IDA and rA using the following equation:\nyA = yh(IDA,rA)\nZ\n· rA (mod p)\n(9.1)\nWe denote this initial key pair as (xA,0, yA,0).\n■\nUser-controlled key pair progression: Node A can update its (pub-\nlic, private) key pair either synchronously or asynchronously. In the\nsynchronous setting, where A uses the key pair (xA,t, yA,t) in time in-\nterval [t ·\u0003T, (t +1)·\u0003T ), node A can choose n random pairs {kA,t ∈\nZ∗\nq, rA,t = αkA,t (mod p)}, where 1 ≤t ≤n, and publishes guaran-\ntees rA,t. Then the private key of node A progresses as follows:\nxA,t = xA,0 · h(IDA, rA,t) + kA,t (mod q)\n(9.2)\n" }, { "page_number": 313, "text": "302\n■\nSecurity in Wireless Mesh Networks\nand the corresponding public keys can be computed according to\nyA,t = yh(IDA,rA,t)\nA,0\n· rA,t (mod p)\n(9.3)\n■\nNon-interactive pairwise key agreement and progression: Pairwise\nshared keys between any two nodes A and B can also be com-\nputed and updated synchronously or asynchronously based on\nAlgorithm 1.\nAlgorithm\n1\nKey agreement\nbetween nodes A and B\nNode A:\nxA,t = xA,0 · h(IDA,r A,t) + kA,t\nyB,t = y\nh(IDB,r B,t)\nB,0\n· r B,t (mod p)\nK A,t = yxA,t\nB,t (mod p)\nK t = h(K A,t)\nNode B:\nxB,t = xB,0 · h(IDB,r B,t) + kB,t\nyA,t = y\nh(IDA,r A,t)\nA,0\n· r A,t (mod p)\nK B,t = yxB,t\nA,t (mod p)\nK t = h(K B,t)\nThe pairwise shared keys obtained by node A and node B are equal\nbecause\nh(K A,t) = h(yxA,t\nB,t (mod p))\n= h(αxA,t xB,t (mod p)) = h(yxB,t\nA,t (mod p)) = h(K B,t)\n(9.4)\nTwo features of SCK are worth pointing out:\n1.\nGiven that N nodes participate in the network and their IDs are\nglobally known, N guarantees are advertised to distribute their pub-\nlic keys, instead of N traditional certificates. The advantage is that,\nunlike a certificate-based approach, such N guarantees can be pub-\nlished and need not be certified (signed) by any centralized author-\nity. This means that the public key of each node can be derived and\nupdated (rekeying) without the aid of an online CA (access to the\nCA is only required at the initial network formation, as previously\ndescribed).\n2.\nGiven that guarantees are correctly received by each node in the\nnetwork, then any two nodes can establish and progress the pair-\nwise key shared between them in a non-interactive manner. Con-\nsequently, without considering the distribution of guarantees, the\ncommunication overhead incurred by key establishment is zero.\n" }, { "page_number": 314, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n303\n9.3\nNon-Interactive Key Agreement and Progression\n9.3.1\nS-NIKAP and A-NIKAP\nSCK is particularly attractive to the design of security protocols for wire-\nless networks because it promises a NIKAP scheme. However, the basic\nprimitives of SCK cannot be applied directly to WMNs. In this section, we\npresent two protocols that implement NIKAP to facilitate security mecha-\nnisms using symmetric cryptographic primitives, and allow NIKAP to be\nconfigurable, depending on whether time synchronization is available to\nwireless nodes in the network.\nFor NIKAP to work correctly, we assume that the guarantees of a node\nare successfully distributed to all nodes participating in the network. To en-\nsure the delivery of nodal guarantees in such an error-prone environment\nas wireless channel, an efficient and reliable broadcasting scheme (for in-\nstance, the reliable broadcasting protocol proposed in [21]) can be used to\nfacilitate the process of guarantee distribution, which tolerates link failures\nand node mobility.\nIn S-NIKAP, two nodes negotiate and update the shared keys between\nthem periodically according to the current time instant and the specified\nsecurity policy. Processes or applications of higher security concern can\nperform the rekeying (key progression) operation at a high rate and those\nof lower security concern at a low rate, accordingly. Therefore, communi-\ncation principals in the network can be distinguished based on different\nsecurity policies, such as roles, service types, or the sensitivity of data.\nAs a result, differentiated security services can be achieved by specifying\nhigh-to-low rekeying rates that correspond to high-to-low security levels.\nThe main limitations of S-NIKAP are the prerequisite of time synchroniza-\ntion and the periodical rekeying at a fixed rate. Though there exist de-\nvices or protocols providing time synchronization for wireless networks,\nit is still not clear if the desired performance can be achieved in such dy-\nnamic and unpredictable environments. Another drawback of S-NIKAP is\nthat the pairwise key is independently updated no matter whether there\nis communication between peer nodes to take place. Therefore, local CPU\ncycles (and therefore battery life) are wasted if the newly generated keys\nare not used within its life cycle. Algorithm 2 presents the specification of\nS-NIKAP.\nAlgorithm\n2\nProtocol S-NIKAP(for any node A)\n1. Node initialization:\nRetrieve the CA’s public key yZ, initial private key xA,0,\ninitial guarantee rA,0, and key progression interval \u0003T\n2. Guarantees distribution : Advertise IDA and randomly selected guarantees rA,t\nwhere 1 ≤t ≤n. (rA,t and IDA can be broadcast over insecure channel)\n" }, { "page_number": 315, "text": "304\n■\nSecurity in Wireless Mesh Networks\n3. Pairwise keys agreement\nand progression : To communicate with node B\nwithin time interval [T0 + t · \u0003T, T0 + (t + 1) · \u0003T ), first update the key shared with\nB to Kt, according to the following procedure:\nxA,t = xA,0 · h(IDA, rA,t) + kA,t\nyB,t = yh(IDB,rB,t)\nB,0\n· rB,t(mod p)\nK A,t = yxA,t\nB,t (mod p)\nKt = h(K A,t)\nIt follows naturally that an asynchronous version of NIKAP is desired in\ncases in which time synchronization is not available or portable nodes can-\nnot afford the cost of progressing keys at high rates. A-NIKAP has the same\nnon-interactive rekeying capability as S-NIKAP does, but requires no time\nsynchronization service from the underlying network. Instead, A-NIKAP\nuses a pseudo-random bit stream to synchronize the rekeying process be-\ntween nodes, of which “1” invokes new key progression and “0” keeps\ntwo nodes using the current key shared between them. According to SCK,\nan initial shared key can be non-interactively established. Therefore, the\npseudo-random bit stream can be generated, encrypted (using the initial\nkey), and securely agreed-upon between nodes sharing the initial key. If\nthe same pseudo-random number generator is used by both ends, to save\nthe bandwidth, only a common seed needs to be exchanged. The pro-\ngression strategy in A-NIKAP can be specified as per-session based, fixed\nnumber of sessions based, or fixed number of packets sent based etc.,\naccording to the given security policies. If the bit-synchronization is lost,\nnodes need to re-establish a new pseudo-random bit stream (by using the\nlast pairwise key working between them, or simply start over). If we count\none bit in the random bit stream equal to one time interval used in S-NIKAP,\nA-NIKAP incurs half of the local CPU cycles than S-NIKAP does, provided\nthat the bit stream is perfectly randomized. Algorithm 3 defines protocol\nA-NIKAP.\nAlgorithm\n3\nProtocol A-NIKAP(for any node A)\n1. Node initialization:\nRetrieve CA’s public key yZ, initial private key xA,0, and\ninitial guarantee rA,0\n2. Guarantees distribution:\nAdvertise IDA and randomly selected guarantees rA,t,\nwhere 1 ≤t ≤n. (rA,t and IDA can be broadcast over insecure channel)\n3. Random bits stream generation\nand exchange:\nTo communicate with node\nB, first generate a random bit stream BITSA and send to B as follows:\nA ⇒B : {IDA, IDB, BITSA,\nhash (IDA, IDB, BITSA, K A,0)}K A,0\nWhere the hashing value hash (·) is used by node B to verify the integrity of BITSA\n" }, { "page_number": 316, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n305\n4. Bit-controlled\nkey progression:\nWhile BITSA is not empty do\nif new session then\n/* Or other triggering events */\nf lag ←pop(BITSA)\nif f lag = 1 then\nupdate the shared key to Kt\nelse\nkeep using the current key Kt−1\n9.3.2\nApplication Scenarios of NIKAP\nThe non-interactive progression capability of NIKAP makes it attractive to\nwireless applications in which shared keys need to be established without\nnegotiation through insecure channels, or need to be updated frequently.\nSuch scenarios include secure ad hoc routing, peer-to-peer communication\nin combat fields, and surveillance systems.\nWhen mechanisms based on symmetric cryptographic algorithms are\nused to secure the routing discovery process in wireless ad hoc networks,\ninteractive key agreement protocols are not suitable, because the topology\nand routes in an ad hoc network are usually unknown when it is first de-\nployed. Consequently, given that there can be no pre-existing routes for\nnodes to communicate with each other, a common broadcast channel must\nbe used for key establishment, which is easy to be exploited by malicious\nusers. In addition, requiring the collaboration among nodes to establish\nshared keys while they are establishing routes to one another cannot be\ndone efficiently. The non-interactive nature of NIKAP allows nodes to se-\ncure the routing process without incurring undue overhead.\nNIKAP can also be used to provide differentiated security services in\nwireless networks. To achieve better security, the keys shared between\nnodes can be updated regularly, and the keys used between different nodes\ncan be rekeyed at different rates based on different security policies, such\nas privilege rankings, roles, and location of the nodes.\nSurveillance systems are often used to gather and upload critical data\nperiodically to a command center from monitoring nodes. The topology\nof a surveillance system is relatively fixed compared with that of a mobile\nad hoc network (MANET), which exposes it to high possibility of being\nidentified and attacked. Therefore, keys used between the command cen-\nter and each monitoring node have to be updated regularly. Moreover, a\npairwise key-based scheme is also preferable to a group key-based scheme,\nto confine the damage caused by key divulgence. In such a case, S-NIKAP\ncan be a good candidate for key establishment because of its periodic,\nnon-interactive key progression capability.\n" }, { "page_number": 317, "text": "306\n■\nSecurity in Wireless Mesh Networks\n9.4\nAd hoc On-demand Secure Routing Protocol\nIn this section, we present the secure Ad hoc On-demand Secure Routing\nprotocol (AOSR), which derives pairwise keys using NIKAP and exploits\nkeyed hash values to authenticate the generic on-demand ad hoc routing.\n9.4.1\nAssumptions\nWe assume that each pair of nodes (node Ni and node N j) in the network\nshares a pairwise secret key Ki, j, which can be achieved by using the key\nagreement protocols described in Section 9.3.1. Whether S-NIKAP or A-\nNIKAP is adopted depends on the availability of time synchronization in\nthe deployed network. We also assume that the MAC address of a node\ncannot be changed once it joins the network. Even though some vendors\nof modern wireless cards do allow a user to change the card’s MAC address,\nwe will see that this simple assumption can be helpful in detecting some\ncomplicated attacks such as wormhole. Moreover, every node must obtain\na certificate signed by the CA, which binds its MAC and ID (can be the IP\naddress of this node), before it joins the network. Note that such certificates\nare used for nodes to verify the authenticity of their neighbors, rather than\nvalidating the routes discovered during the process of route discovery. A\nnode presents its certificate to each node that it meets for the first time,\nand two nodes can communicate with its neighbor nodes only if their\ncertificates have been mutually verified. The approach used to authenticate\nand maintain neighbor-node information is presented in [5], and as such is\nomitted here due to space limitations. To be clear, the notation used in the\nrest of the chapter is summarized in Table 9.1.\n9.4.2\nRoute Discovery\nAOSR consists of route request initialization, route request forwarding, route\nrequest checking at the destination D, and the symmetric route reply initial-\nization, route reply forwarding, and route reply checking at the source S.\nThe message flow of the route discovery of AOSR is illustrated in Figure 9.1.\n9.4.2.1\nRoute Request Initialization\nSource S generates the following route request RREQ and broadcasts to its\nneighboring nodes, when S wants to communicate with node D, but has\nno active route maintained for D at that point.\nRREQ = {RREQ, S, D, QNum, HC, {NodeList}, QMACs,d}\n(9.5)\nbecause no node has been traversed by RREQ at the source S, HC = 0 and\n{NodeList} = {Null}. QMACs,d = Hash(CORE, HC, {NodeList}, Ks,d) is the\n" }, { "page_number": 318, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n307\nTable 9.1\nNotation Used in This Chapter\nName\nMeaning\nS, D, Ni\nNode IDs, particularly, S = source, D = Destination\nRREQ\nThe type identifier for a route request RREQ\nRREP\nThe type identifier for a route reply RREP\nRERR\nThe type identifier for a route error report RERR\nQNum\nThe route request ID, a randomly generated number\nRNum\nThe route reply ID, and RNum = QNum + 1 for the same round of\nroute discovery\nHCi→j\nThe hop count from node Ni to Nj\nQMAC\nThe k-MAC2 used in RREQ\nRMAC\nThe k-MAC used in RREP\nE MAC\nThe k-MAC used in RERR\nKi, j\nThe key shared between nodes Ni and Nj, thus Ki, j = K j,i\n{NodeList}\nRecords the accumulated intermediate nodes traversed by messages\nRREQ, RREP, or RERR. For clarity, they are increasingly numbered\nfrom S to D, i.e., {S, N1, N2...Ni...D}\nr Ti→j\nThe route from node Ni to node Nj\nk-MAC which will be further processed by intermediate nodes and used\nby the destination D to verify the integrity of RREQ and the validity of the\npath recorded by {NodeList}. Parameter\nCORE = Hash(RREQ, S, D, QNum, Ks,d)\n(9.6)\nserves as a credential of S to assure D that the RREQ is really originated\nfrom S and its immutable fields are integral during the propagation.\nS\nD\nK\nJ\nI\nL\nRREPD,L\nRREPI,S\nRREPJ,I\nRREPK,J\nRREPL,K\nRREQS,I\nRREQI,J\nRREQJ,K\nRREQK,L\nRREQL,D\nD\nRREQ: S\nD\nRREP: S\nFigure 9.1\nRoute discovery between source S and destination D.\n2 In our discussion, k-MAC refers to keyed-message authentication code (a keyed hash\nvalue), while MAC refers to media access control unless specified otherwise.\n" }, { "page_number": 319, "text": "308\n■\nSecurity in Wireless Mesh Networks\n9.4.2.2\nRoute Request Forwarding\nAn RREQ received by an intermediate node Ni is processed and further\nbroadcast only if it has never been seen (the ID of node S and the randomly\ngenerated QNum uniquely identify the current route discovery initialized by\nS). Because {NodeList} records the nodes that have been traversed before\nthe RREQ is received at Ni, Ni increases HC by one and appends the ID\nof the upstream node Ni−1 into {NodeList}, and updates QMAC as follows:\nQMACi,d = Hash(QMACi−1,d, HC, {NodeList}, Ki,d)\n(9.7)\nA reverse forwarding entry is also established at Ni, which is used to relay\nthe corresponding RREP back to source S.\n9.4.2.3\nChecking RREQ at Destination D\nFigure 9.2 shows the procedure conducted by destination D to authenticate\nthe validity of the path reported by RREQ. Basically, D repeats the com-\nputation executed by each intermediate node traversed by RREQ, which is\nrecorded in field {NodeList}, using the shared keys maintained by D itself.\nObviously, the number of hashing that D needs to perform equals HC, the\nnumber of nodes traversed by the RREQ.\nIf such a verification is successful, D can be assured that the RREQ was\nreally originated from S, each node listed in {NodeList} actually participated\nin the forwarding of RREQ, and the distance between S and D is equal to\nHCs→d.\nThe route reply initialization, reverse forwarding of route reply, and\nchecking RREP at the source S are basically symmetric to that of RREQ,\nand as such are omitted for brevity. Note that AOSR forwards traffic on\na hop-by-hop basis, and each intermediate node relaying an RREP also\nestablishes the forwarding entry for the requested destination D, which is\nused to route succeeding data packets.\n9.4.3\nRoute Maintenance\nA route error message (RERR) is generated and unicast back to source S\nif an intermediate node Ni finds the downstream link of an active route\nis broken (Figure 9.3). Before accepting an RERR, S must make sure that\n(1) the node generating the RERR belongs to the path for the destination,\nand (2) the node reporting link failure should actually be there when it is\nreporting the link failure. The process of sending back an RERR from node\nNi is similar to that of originating a route reply from Ni to the source S.\n" }, { "page_number": 320, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n309\nNo\nDrop RREQ\nYes\nNo\nDrop RREQ\nYes\nYes\nNo\nAccept RREQ\nQMACTemp = Hash(CORE, 0, {NodeList}, Keyd,s);\nCORE = Hash(RREQ, S, D, QNum, Keyd,s);\n{NodeList} = {Null}; \nQMACHC,d = QMACTemp\n{NodeList} = {NodeList, Nj}; //N0 = S \nQMACTemp = Hash(QMACTemp, j, {NodeList}, Keyd,j);\nj = j + 1;\nSeen RREQ ?\nj = 0 \nj < HC\nFigure 9.2\nCheck RREQ at destination D.\nS\nD\nNi−1\nNi + 1\nNi\nRERRi,i–1\nRERR\nCORE = Hash(RERR, Ni, S, RNum, Ki,s)\nEMACi,s = Hash(CORE, 0, {Null}, Ki,s)\nRERRi,i−1 = {RERR, Ni, S, RNum, 0, {Null}, EMACi,s}\nFigure 9.3\nNi generates RERR when the downstream link fails.\n" }, { "page_number": 321, "text": "310\n■\nSecurity in Wireless Mesh Networks\nTherefore, here we only describe the main differences. An RERR has a\nformat similar to that of an RREP, except the type identifier RERR and the\ninitialization of CORE, which is calculated as follows:\nCORE = Hash(RERR, Ni, S, D, RNum, Ki,s)\n(9.8)\nEach intermediate node in the reverse path to the source only processes\nand back-forwards an RERR received from its successor used for destination\nD, which ensures that no node rather than Ni can initialize an RERR, and\nnode Ni is still in the path for D when reporting the link failure. When\nsource S receives the RERR, it invokes a verification procedure similar to\nthat of RREP. The only difference is the initial value of CORE, which is\ncalculated by\nCORE = Hash(RERR, Ni, S, D, RNum, Ks,i)\n(9.9)\nwhere rather than Ki,s of node Ni, the pairwise key Ks,i maintained at S is\nused.\n9.5\nSecurity Analysis\nThe attacks to an ad hoc network can be classified into external attacks\nand internal attacks based on the information acquired by the attackers.\nExternal attacks are launched by malicious users who do not have the\ncryptographic credentials (e.g., the keys required by the cryptographic al-\ngorithms being used) that are needed to participate in the route discovery.\nOn the other hand, internal attacks are originated by attackers who have\nbroken into legitimate nodes, and as such have access to cryptographic\nkeys owned by the compromised nodes. As a result, internal attacks are\nfar more difficult to detect and not as defensible as external attacks. For a\ngood description of potential attacks to ad hoc routing, the reader can refer\nto [6,7]. Figure 9.4 depicts the network topology and notation used for our\nanalysis. In the following, we only consider RREQ because the processing\nof RREP is symmetric.\nIn AOSR, a route request RREQ consists of immutable fields RREQ,\nQNum, S, D, and mutable fields QMAC, HC, and {NodeList}. As to im-\nmutable parts, they are protected by the one-way hash value CORE, which\nhas RREQ, S, D, QNum, and Ks,d as the input. No node can impersonate\nthe initiator S to fabricate RREQ due to the lack of key Ks,d known only\nto S and D. Any modification on such fields can be easily detected by\ndestination D, because the QMAC carried in the RREQ cannot match what\nD recalculates based on {NodeList}.\nMutable fields {HC, {NodeList}, QMAC} are modified by intermediate\nnodes when the RREQ propagates to D. In AOSR, the authenticity of\n" }, { "page_number": 322, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n311\nWormhole tunnel\nNon-existent links\nR : Node removed by A1\nI : Node forged by A2\nV1,V2 : Nodes forged by W1,W2\nW1,W2 : Two nodes ( form a wormhole)\nA1, A2 : Attackers\nD : Destination\nS : Source\nS\nD\n3\nR\n1\nW1\nV1\nV2\n2\nI\nW2\nA1\nA2\nFigure 9.4\nNetwork topology for security analysis.\nHC, {NodeList}, and QMAC is guaranteed by integrating HC and {NodeList}\ninto the computation of QMAC, in such a way that no node can be added\ninto {NodeList} by the downstream node, unless it has actually forwarded\nan RREQ; and no node can be maliciously removed from {NodeList}, un-\nless it is not used for routing traffic for D. For instance, let us assume\nthat attacker A1 attempts to remove node R from {NodeList} and decrease\nHC by one. When receiving the RREQ, D recomputes QMAC accord-\ning to the nodes listed in {NodeList}. Because the hashing executed by\nR, i.e., QMACr,d, has been omitted, D cannot have a match with the\nreceived QMAC. The reason is that hashing operation is one-way only,\nand there is no way for A1 to reverse the computation of QMACr,d. An-\nother possible attack is for attacker A2 to insert a non-existent node I\ninto {NodeList} and increase HC by one. To achieve this, A2 needs to\nperform one more hashing that requires Ki,d as the input, which is im-\npossible because Ki,d is only known to I and D. For the same reason,\nA2 cannot impersonate another node (spoofing) and make itself appear\non {NodeList}.\nA wormhole is a special attack that is notoriously difficult to detect and\ndefend against. Wormholes usually consist of two or more nodes working\ncollusively, picking up packets at one point of the network, tunneling them\nthrough a special channel, then releasing them at another point far away.\nThe goal is to mislead the nodes near the releasing point to believe that\nthe tunneled packets are transmitted by a nearby node. A demonstrative\nscenario of wormhole attacks is shown in Figure 9.5. Wormholes are a big\n" }, { "page_number": 323, "text": "312\n■\nSecurity in Wireless Mesh Networks\nS\nD\nN1\nN2\nPacket\nPacket\nPacket\nImplicit tunnel between N1 and N2\nFigure 9.5\nIllustration of wormhole attack. Shaded nodes are attackers and white\nnodes are legal nodes.\nthreat to ad hoc routing, largely because wrong topology information is\nlearned by the nodes near the releasing point. As a result, data packets are\nmore likely to be diverted into the tunnel, in which attackers can conduct\nvaried malicious operations, such as dropping data packets (black hole\nattack), modifying packet contents, or performing traffic analysis.\nWormholes can be further classified based on the type of end nodes\nforming the tunnel. For external attackers (without valid keys or certifi-\ncates), they need to make themselves invisible due to the lack of required\nkeys to participate in the routing process. Therefore, what they actually\nperform is passing packets through the tunnel without any modification.\nOn the other hand, internal attackers can “legally” participate in the routing\nprocess, and as such manipulate the intercepted packets with many more\npossibilities.\nThe chained k-MAC values computed by all intermediate nodes during\nthe route discovery, together with the authenticated neighbor information\nprovided by the neighbor maintenance scheme, enable AOSR to detect a\nwormhole and varied attacks derived from it. As an example, let us assume\nthat nodes W1 and W2 in Figure 9.4 are two adversaries who have formed\na tunnel T ulw1↔w2. First, they can refuse to forward RREQ, but this is not\nattractive because this actually excludes them from the route discovery.\nSecond, they can attempt to modify HC or {NodeList}, but this can be\ndetected when destination D checks the QMAC carried by RREQ. They can\nalso insert some non-existent nodes, like V1, V2, into {NodeList}, but this\ncannot succeed due to the lack of shared keys Kv1,d and Kv2,d.\nPackets tunneled by external attackers can be detected because the MAC\naddress of the outsider cannot match any ID maintained by the neighbor\nlist at the receiving node near the releasing point (or does not exist at\nall). This can be done because a node’s MAC address cannot be changed,\nany binding of a MAC address and an ID on the neighbor list has been\nauthenticated, and the MAC address of a packet is always in cleartext. For\ninstance, assume again that the nodes W1 and W2 in Figure 9.4 are two\n" }, { "page_number": 324, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n313\nexternal attackers and form a tunnel T ulw1↔w2, and w1 or w2 is tunneling\na packet from node 2 to node D. This packet cannot be accepted because\nthe MAC address shown in the packet (the MAC address of W2) does not\nmatch the MAC address of node 2 maintained by node D (or there is no\nneighbor entry maintained for node 2 at all).\nThe only variation of wormhole attacks that AOSR cannot detect takes\nplace when the end node at the releasing point is an internal attacker to\nthe network, and owns all the required cryptographic keys or certificates.\nTo date, there is still no effective way to detect this kind of wormhole\nattack. Though there are other approaches to defending against wormhole\nattacks [8], time synchronization must be made available to each node for\nthe proposed packet leashes to work. On the other hand, binding on an\nunalterable MAC address with a nodal identifier is simple to implement and\nprovides almost the same defensive results as packet leashes.\n9.6\nPerformance Evaluation\nWe implement AOSR in NS2 [9], which can act as the centralized authority\nat the network formation and provide time synchronization in the course\nof simulation. Therefore, S-NIKAP is used to serve the purpose of key\nestablishment among mobile nodes. The hash function (used for the com-\nputation of k-MAC) and the digital signing function (used by the neighbor\nmaintenance scheme) in our simulation are MD5 (128 bits) and RSA (1024\nbits), respectively. In this way, we take into account the cost and delay\ncaused by the cryptographic operations performed by AOSR, in addition to\nthe overhead incurred by processing control messages. The simulation pa-\nrameters are summarized in Table 9.2, and used throughout the following,\nunless specified otherwise.\nFive metrics are used to evaluate the performance of AOSR:\n1.\nPacket delivery ratio (PDR) is the total number of CBR packets re-\nceived, over the total number of CBR packets originated, averaged\nover all nodes in the network.\n2.\nEnd-to-end packet delay is the average elapsed time between a CBR\npacket passing to the routing layer and that packet being received\nat the destination node, averaged over all received packets.\n3.\nRoute discovery delay is the average time it takes for the source\nnode to find a route for the requested destination.\n4.\nNormalized routing overhead is the total routing messages origi-\nnated and forwarded over the total number of CBR packets received,\naveraged over all nodes.\n5.\nAverage route length is the average length (hops) of the routes used\nto forward data packets, averaged over all routes discovered.\n" }, { "page_number": 325, "text": "314\n■\nSecurity in Wireless Mesh Networks\nTable 9.2\nSimulation Parameters\nParameter\nValue\nSimulator\nNS2 [9]\nTopology\n30 nodes, 1000 m × 250 m field\nNode placement\nUniformly distributed\nPropagation model\nTwo-ray propagation\nMAC protocol\n802.11 DCF\nTransmission range\n250 m\nLink bandwidth\n2 × 106 bps\nTraffic pattern\n15 constant bit rate (CBR) flows with randomly\nchosen source and destination, two packets per\nsecond, and with a payload size of 512 bytes. Each\nflow starts randomly within 50 seconds after the\nsimulation is launched, and the lasting time varies\nbetween 100 ∼200 seconds\nMobility model\nRandom way-point model with Vmin = 0 and\nVmax = 15 mps\nSimulation time\n300 seconds\n# of trials with random seeds\n5\nFigure 9.6 and Figure 9.7 demonstrate the performance comparison be-\ntween AOSR using S-NIKAP and the AODV protocol [10]. When there is\nno attack occurring in the network, the normalized routing overhead of\nAOSR, as shown in Figure 9.6(b), is almost the same as that of AODV. The\nreason is intuitive: establishing shared keys using NIKAP does not need the\nnegotiation between nodes or between the nodes and an online CA. In our\nsimulation, the key progression interval is set to five seconds, and in prac-\ntice, this is adjustable according to the processing power of mobile nodes,\nor the given security policy. Because shared keys between nodes need to\nbe updated at a fixed rate, we expect that the time it takes for AOSR to\ndiscover routes should be longer than that of AODV. Fortunately, as shown\nin Figure 9.7(a), the average routing delay caused by key progression, mea-\nsured over all nodes, is only 2 ∼5 milliseconds more than that of AODV,\nwhich is an acceptable increase of 5∼12 percent. This indicates that NIKAP\nefficiently supports the security mechanisms used by the route-discovery\nprocess of AOSR without incurring significant routing delay. The average\nroute length of AOSR is a little shorter than that of AODV, as shown in\nFigure 9.7(c). The reason is that AOSR requires all route requests to reach\nthe destination, while AODV allows intermediate nodes to reply to an RREQ\nif they cache an active route, which may not be the shortest at that moment.\nThis also explains why the packet delivery delay of AOSR is shorter than\nthat of AODV, as shown in Figure 9.7(b).\n" }, { "page_number": 326, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n315\n0\n100\n200\n300\n0.962\n0.964\n0.966\n0.968\n0.97\n0.972\n0.974\n0.976\n0.978\n0.98\nPause time (seconds)\n(a) Packet delivery ratio PDR\nAODV\nAOSR\n0\n100\n200\n300\n0.8\n0.85\n0.9\n0.95\n1\n1.05\n1.1\n1.15\n1.2\n1.25\nPause time (seconds)\n(b) Normalized routing overhead\nAODV\nAOSR\nFigure 9.6\nPDR and routing overhead comparisons without attackers.\n0\n100\n200\n300\n0.032\n0.033\n0.034\n0.035\n0.036\n0.037\n0.038\nPause time (seconds)\n0\n100\n200\n300\n0.015\n0.02\n0.025\n0.03\n0.035\n0.04\nPause time (seconds)\n0\n100\n200\n300\n2.51\n2.52\n2.53\n2.54\n2.55\n2.56\n2.57\n2.58\nPause time (seconds)\nAODV\nAOSR\nAODV\nAOSR\nAODV\nAOSR\n(a) Routing delay (seconds)\n(b) Packet delay (seconds)\n(c) Route length (hops)\nFigure 9.7\nDelay and route-length comparisons without attackers.\n" }, { "page_number": 327, "text": "316\n■\nSecurity in Wireless Mesh Networks\n0\n100\n200\n300\n0.2\n0.3\n0.4\n0.5\n0.6\n0.7\n0.8\n(a) Packet delivery rate PDR\nPause time (seconds)\n0\n100\n200\n300\n0\n0.5\n1\n1.5\n2\n2.5\n3\n(b) Normalized routing overhead\nPause time (seconds)\nAODV−30\nAODV−60\nAOSR−30\nAOSR−60\nAODV−30\nAODV−60\nAOSR−30\nAOSR−60\nFigure 9.8\nPDR and routing overhead comparisons with attackers.\nFigure 9.8 and Figure 9.9 present the simulation results when 30 and 60\npercent of the nodes in the network are compromised, and fabricate fake\nroute replies to route requests by claiming that they are zero hop away\nfrom the specified destination node, in hopes that the querying source\nnode is willing to send its succeeding data packets to them. After that, a\ncompromised node simply drops all the data packets received (black hole\nattack).\nThe packet delivery ratio of AODV decreases drastically, as shown in\nFigure 9.8(a), given that most of the packets are sent to the compromised\nnodes, which discard them silently. The average route length of AODV is\nmuch shorter than when there is no malicious node in the network, as\nshown in Figure 9.9(c). The reason is that a compromised node is likely to\nreceive and reply to the route requests for the specified destination earlier\nthan the destination itself or other nodes having an active route. This also\nindicates that most of the successful packets are delivered within one or\ntwo hops away from the source.\nOn the other hand, as shown in Figure 9.8(a), AOSR is still able to sus-\ntain over 62 percent packet delivery ratios for all pause time configurations,\neven when 60 percent of the nodes are compromised. This is achieved at\nthe cost of more routing time to find a route, longer end-to-end packet de-\nlay, and higher routing overhead, as shown in Figure 9.9(a), Figure 9.9(b),\n" }, { "page_number": 328, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n317\n0\n100\n200\n300\n0.025\n0.03\n0.035\n0.04\n0.045\nPause time (seconds)\n(a) Routing delay (seconds)\nAODV−30\nAODV−60\nAOSR−30\nAOSR−60\n0\n100\n200\n300\n0\n0.1\n0.2\n0.3\n0.4\n0.5\nPause time (seconds)\n(b) Packet delay (seconds)\n0\n100\n200\n300\n1\n1.5\n2\n2.5\nPause time (seconds)\n(c) Route length (hops)\nAODV−30\nAODV−60\nAOSR−30\nAOSR−60\nAODV−30\nAODV−60\nAOSR−30\nAOSR−60\nFigure 9.9\nDelay and route-length comparisons with attackers.\nand Figure 9.8(b), respectively. Lastly, nodes running AOSR cannot be mis-\nled by compromised nodes declaring better reachability for the requested\ndestination,3 and as such are able to find a route to the destination if there\nis one. Consequently, the average length of routes discovered by AOSR is\nlonger than that of AODV, as shown in Figure 9.9(c).\n9.7\nRelated Work and Open Issues\nExisting key distribution protocols for wireless networks generally assume\nthe existence of an online CA. To alleviate the risk caused by the single\npoint of failure, threshold cryptography replaces the CA by a subset of\nnodes that share and provide the functionality of the CA contributorily [2].\nHowever, this approach cannot completely eliminate the reliance on the\nfunctioning of an online CA, which is still of major interest to attackers.\n3 AOSR detects the misbehavior of malicious nodes when the verification of RREQ or\nRREP fails.\n" }, { "page_number": 329, "text": "318\n■\nSecurity in Wireless Mesh Networks\nThe alternative use of multiple mobile mini-CAs requires nodes to contact\nup to a certain number of mini-CAs before they can obtain the desired\nkeys. Therefore, we have reason to argue that, in highly dynamic scenarios\nsuch as WMNs, the responsiveness of deploying multiple mini-CAs could\nbe worse than schemes based on a single CA. Key distribution protocols\nusing ID-based cryptography [11], or the combination of threshold and ID-\nbased cryptography [12], have the same advantage as SCK because IDs\n(publishable) are used to obtain the corresponding public keys of nodes,\ninstead of using a certificate to bind the ID and its public key. However,\nonline CA services must exist for such protocols to work, which has the\nsame limitations of protocols based on threshold cryptography.\nAnother approach to key agreement for wireless networks is to combine\nthreshold secret sharing and probabilistic key sharing [13]. The basic idea is\nto split the shared secret between a source–target pair into several pieces,\nand propagate them toward the target in such a way that the target node has\na high probability to recover the splitted secret based on the secret pieces it\nreceives. However, the overhead incurred by sending multiple secret pieces\ntoward each target node can be high due to network dynamics. Moreover,\nif a required number of secret pieces do not reach the target, the original\nsecret cannot be recovered.\nGroup key agreement protocols [14,15] are very different from S-NIKAP\nand A-NIKAP. In group key agreement, a shared key needs to be distributed\namong all possible nodes belonging to a multicast or many-to-many-cast\ngroup, while S-NIKAP and A-NIKAP only consider the key agreement be-\ntween two nodes. The storage complexity of a system using group keys\nis obviously lower than that of a system using pairwise keys. However,\nin group communication, the cost of rekeying operation caused by nodes\nleaving or joining a group, network partition, or merging can be consider-\nably high. The reason is that, whenever the group membership changes, a\nnew group key must be re-established among all group members; other-\nwise, the subsequent communication within the group becomes insecure\ndue to the possibility of key divulgence. Another drawback of a system\nusing group keys is that the compromise of a group key can jeopardize the\ncommunication confidentiality of the entire group, while the compromise\nof a pairwise key only affects the pair of nodes using the shared key. In\npractice, whether to use a pairwise key scheme or a group key scheme\nshould be decided according to the application scenario and the security\npolicy.\nFuture design of key management schemes needs to carefully con-\nsider the unique characteristics of wireless networks, i.e., volatile topol-\nogy, collision-prone transmission channel, and stringent resources of the\nwireless nodes. Given that no centralized administration exists, a practical\nkey management scheme must also be fully distributed and self-organizing.\n" }, { "page_number": 330, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n319\nThough threshold cryptograph-based approaches [2] divide the centralized\nauthority into a subset of the nodes to improve the service availability and\nfault tolerance, the inherent idea of central administration limits its appli-\ncability to ad hoc networks, and also makes CA-capable nodes the major\ninterests to malicious attackers. A possible modification to threshold cryp-\ntograph is to allocate each designated mini-CA more than one share of the\nCA’s secret, such that the probability of successfully issuing a certificate can\nbe increased [19].\nThe key pre-distribution (KPD) [18] scheme has been demonstrated to\nbe a promising approach for symmetric key establishment for wireless sce-\nnarios. Given that N is the set of nodes in the network, each node in N\nis first pre-loaded a set of keys chosen from a pre-established key pool.\nThen any sub-group of nodes Ni ⊂N can establish a common key shared\namong them that is unknown to nodes outside Ni. KPD systems have been\nbelieved to be the only practical approach for truly ad hoc scenarios. The\nmajor limitations of KPD are that (1) the success of key establishment is\nprobabilistic guaranteed and (2) the overhead of key pre-distribution can\nbe expensive. An interesting research topic is how to achieve the same\nkey establishment results as that of KPD, but with a deterministic success\nguarantee.\nSignature aggregation [20] is another effective approach to reducing the\nsize of certificate chains by aggregating all certificates in the chains into\na single short signature, as such saves the scarce bandwidth of nodes in\nWMNs. The basic idea of signature aggregation is that, given that N distinct\nmessages are signed by N distinct users, it is possible to aggregate the\nresulting signatures into a single signature in such a way that a verifier of\nthe aggregated signature can be convinced that each user indeed signed its\nmessage. It is an interesting research topic whether such an approach can\nbe utilized for key management for WMNs, especially in the case of group\nkey establishment, to reduce the overhead incurred by group-key creation\nand rekeying.\n9.8\nConclusion\nWe proposed S-NIKAP and A-NIKAP, two key agreement protocols that\nachieve non-interactive key establishment and, if needed, the succeeding\nkey progression (rekeying process). NIKAP needs the aid of a centralized\nauthority only at the initial network formation, which is better than other\napproaches relying on online CA services. Our work using NIKAP for secure\nad hoc routing shows that NIKAP bootstraps key establishment in ad hoc\nnetworks efficiently, and is promising for other resource-constrained ad hoc\nscenarios where frequent and non-interactive key rekeying are desired.\n" }, { "page_number": 331, "text": "320\n■\nSecurity in Wireless Mesh Networks\nReferences\n[1]\nJ. Kohl and B. Neuman, The Kerberos Network Authentication Service (V5),\nRFC 1510, September 1993.\n[2]\nL. Zhou and Z.J. Haas, Securing ad hoc networks, IEEE Network, special\nissue on network security, Vol. 13, No. 6, pp. 24–30, 1999.\n[3]\nW. Diffie and E. Hellman, New directions in cryptography, IEEE Tran.\nInform. Theory, Vol. 22, pp. 644–654, November 1976.\n[4]\nH. Petersen and P. Horster, Self-Certified Keys—Concepts and Applica-\ntions, 3rd Conference of Communications and Multimedia Security, Athens,\nSeptember 22–23, 1997.\n[5]\nZ. Li and J.J. Garcia-Luna-Aceves, Enhancing the Security of On-demand\nRouting in Ad hoc Networks, 4th International Conference on Ad-hoc Net-\nworks and Wireless (AdhocNow ’2005), LNCS 3738, pp. 164–177, Cancun,\nMexico, October 6–8, 2005.\n[6]\nK. Sanzgiri, B. Dahill, B.N. Levine, E. Royer, and C. Shields, A Secure Rout-\ning Protocol for Ad hoc Networks, 10th Conference on Network Protocols\n(ICNP), 2002.\n[7]\nY. Hu, A. Perrig, and D. Johnson, Ariadne: A Secure On-demand Routing\nProtocol for Ad hoc Networks, 8th ACM International Conference on Mobile\nComputing and Networking (MobiCom), September 2002.\n[8]\nY. Hu, A. Perrig, and D. Johnson, Packet Leashes: A Defense against Worm-\nhole Attacks in Wireless Networks, IEEE INFOCOM, San Francisco, March\n30–April 3, 2003.\n[9]\nNS2, The Network Simulator, http://www.isi.edu/nsnam/ns/.\n[10]\nC. Perkins, E. Royer, and S. Das, Ad hoc On Demand Distance Vector\n(AODV) Routing, RFC 3561 (Experimental), July 2003.\n[11]\nD. Boneh and M. Franklin, Identity Based Encryption from the Weil Pairing,\nCrypto ’2001, LNCS 2139, pp 213–229, 2001.\n[12]\nA. Khalili, J. Katz, and W. Arbaugh, Towards Secure Key Distribution in\nTruly Ad-hoc Networks, IEEE Workshop on Security and Assurance in Ad\nhoc Networks, Orlando, January 28, 2003.\n[13]\nS. Zhu, S. Xu, S. Setia, and S. Jajodia, Establishing Pairwise Keys for Secure\nCommunication in Ad hoc Networks: A Probabilistic Approach, 11th IEEE\nInternational Conference on Network Protocols (ICNP), Washington, DC,\n2003.\n[14]\nY. Amir, Y. Kim, C. Nita-Rotaru, and G. Tzudik, On the Performance of\nGroup Key Agreement Protocols, 22nd IEEE International Conference on\nDistributed Computing Systems (ICDCS), Vienna, Austria, July 2–5, 2002.\n[15]\nH. Chan, A. Perrig, and D. Song, Random Key Predistribution Schemes for\nSensor Network, IEEE Symposium on Research in Security and Privacy,\npp. 197–213, 2003.\n[16]\nC. Castelluccia, N. Saxena, and J.H. Yi, Self-Configurable Key Pre-\ndistribution in Mobile Ad-hoc Networks, IFIP Networking Conference,\nLNCS 3462, pp. 1083–1095, Waterloo, Canada, May 2005.\n" }, { "page_number": 332, "text": "Non-Interactive Key Establishment in Wireless Mesh Networks\n■\n321\n[17]\nS. Capkun, L. Buttyan, and J.P. Hubaux, Self-organized public-key manage-\nment for mobile ad hoc networks, IEEE Transactions on Mobile Computing,\nVol. 2, No. 1, pp. 52–64, 2003.\n[18]\nA.C.-F. Chan, Distributed Symmetric Key Management for Mobile Ad hoc\nNetworks, IEEE INFOCOM, Hong Kong, March 7–11, 2004.\n[19]\nD. Joshi, K. Namuduri, and R. Pendse, Secure, redundant and fully dis-\ntributed key management scheme for mobile ad hoc networks: An anal-\nysis, EURASIP Journal on Wireless Communications and Networking, pp.\n579–589, 2005(4).\n[20]\nD. Boneh, C. Gentry, H. Shacham, and B. Lynn, Aggregate and Verifiably\nEncrypted Signatures from Bilinear Maps, EuroCrypt ’03, LNCS 2656, pp.\n416–432, 2003.\n[21]\nE. Pagani, Providing reliable and fault tolerant broadcast delivery in mobile\nad-hoc networks, Mob. Netw. Appl., Vol. 4, No. 3, pp. 175–192, 1999.\n" }, { "page_number": 333, "text": "" }, { "page_number": 334, "text": "Chapter 10\nKey Management in\nWireless Mesh Networks\nManel Guerrero Zapata\nContents\n10.1\nIntroduction ........................................................ 324\n10.2\nRelated Work ....................................................... 324\n10.3\nPlaying without a Referee.......................................... 326\n10.3.1\nSymmetric versus Asymmetric Cryptography ............ 326\n10.3.2\nObscurity- and Tamper-Resistant Devices ............... 327\n10.3.3\nMisbehaving Detection Schemes ......................... 327\n10.4\nThe Concept of Identity............................................ 327\n10.4.1\nIdentity in a Place without Authorities................... 328\n10.4.2\nMAC Addresses Are Not Unique Identifiers ............. 328\n10.4.3\nWhat Identifies Me? ....................................... 329\n10.5\nDynamically Generated IP Addresses ............................. 330\n10.5.1\nSAKM IP Address Generation ............................ 330\n10.5.2\nSAKM Message Fields .................................... 331\n10.6\nDuplicated Address Detection ..................................... 332\n10.6.1\nDuplicated IP Address Detection for SAKM ............. 333\n10.6.2\nNetwork Leaders.......................................... 334\n10.7\nDelayed Verification of Signatures ................................. 334\n10.7.1\nRevised Security Requirements........................... 334\n10.7.2\nAchieving Delayed Import Authorization ................ 335\n10.7.3\nSAKM with Delayed Verification ......................... 336\n10.8\nSAKM Encoding of Public Keys and Signatures ................... 337\n10.9\nSAKM Duplicate Address Detection Messages .................... 337\n323\n" }, { "page_number": 335, "text": "324\n■\nSecurity in Wireless Mesh Networks\n10.10 Simulation Results .................................................. 340\n10.11 Open Issues ........................................................ 344\nReferences................................................................. 344\nIn wireless mesh networks (WMN), nodes use the air to communicate, so a\nlot of nodes might hear what a node transmits and there are messages that\nare lost due to collisions. The concept of servers has to be modified: there\nis no guarantee that a node will be able to reach another node, so things\nlike DNS servers, certification authorities (CAs), and other entities that are\nassumed to be found in fixed networks cannot be used here.\nIn a network where the existence of central servers cannot be expected,\nnodes need to be able to communicate without the risk of malicious nodes\nimpersonating the entities they want to communicate with. In a network\nwhere everybody is anonymous, identity and trust need to be redefined.\nIn addition, if the security protocols that are used in these kind of net-\nworks are based in mechanisms that require asymmetric cryptography, the\ntask of having secure routing protocols for such kind of networks will not\nbe completed without an specific key management scheme.\nIn this chapter, we analyze the problems that arise when designing a\nkey management scheme for WMNs. We will use that analysis to design\nSAKM (Simple Ad hoc Key Management), a key management system that\nallows the nodes of an ad hoc network to use asymmetric cryptography\nwith zero configuration, intended to be applied to wireless network routing\nprotocols that provide security features that require the use of asymmetric\ncryptography (like SAODV). Finally, through simulation results, we will\nshow what kind of cryptographic algorithms are more suitable for SAKM\nand for key management in WMNs in general.\n10.1\nIntroduction\nCurrently, there are several secure routing protocols and applications for\nWMN that use symmetric or asymmetric keys without providing a key man-\nagement scheme to distribute them. Some of them argue that a CA can\nbe placed as a special fixed node in the WMN. Nevertheless, this is not\nfeasible if some client nodes are not directly connected to the WMN back-\nbone. In addition, that requires that client nodes need to register to that\nCA. Therefore, there is a need for key management schemes for WMNs\nthat can operate without the help of the WMN backbone, and that allow\nincorporation of new nodes transparently.\n" }, { "page_number": 336, "text": "Key Management in Wireless Mesh Networks\n■\n325\n10.2\nRelated Work\nIn their paper on securing ad hoc networks [28], Zhou and Haas primarily\ndiscuss key management. They devote a section to secure routing, but\nessentially conclude that “nodes can protect routing information in the\nsame way they protect data traffic.” They also observe that denial-of-service\nattacks against routing will be treated as damage and routed around.\nA couple of papers [19,20] have proposed a solution to solve the “ad-\ndress ownership” problem in the context of Mobile IP. It consists in picking\na key pair and mapping the public key to a tentative address in some de-\nterministic way. These ideas can be adapted to the context of WMNs to\nprovide an appropriate key management scheme.\nThe following proposals use symmetric cryptography, and are mainly\ntargeting sensor networks. All of them either assume that there are no\nmalicious nodes, that nodes do not move after deployment, or that no new\nnodes will be added after deployment.\nThe paper about secure pebblenets [4] proposes deploying the same\nsecret key on all nodes to provide group authentication. It has a method\nto select clusterheads to perform the key management. Nevertheless, it\nassumes that there are no malicious nodes and requires nodes to have a\ntamper-resistant storage.\nEschenauer and Gligor [8] propose a scheme that uses a random pre-\ndistribution of secret keys. Each sensor node receives a random subset of\nkeys from a large key pool before deployment. Then, to agree on which\nkey they will use to communicate, two nodes try to find one common key\nwithin their subsets that they can use as their shared secret key. Clearly,\nits main drawback is the requirement of pre-distribution that will not allow\nnew nodes to connect to the network in an ad hoc manner.\nSPINS [22] is a protocol in which sensor networks are formed around\na base station. The base station helps every pair of nodes that need to\ncommunicate in a secure manner to do so. Nevertheless, compromising the\nbase station renders the whole network useless. In addition, each sensor\nnode gets a secret shared with the base station and needs to be able to\ncommunicate with the base station before establishing a communication.\nDu et al. [7] study the problem of random key distribution for networks\nin which there is the knowledge of how the sensor nodes are going to\nbe deployed, which, of course, simplifies a lot the problems of the key\ndistribution. But, it also limits greatly its applicability.\nAnother proposal for static networks is presented in [16], where the main\nidea is that sensor nodes can be deployed with a large amount of keys from\nthe pool of possible keys and, once deployed, decide which keys they keep\naccording to their location and discard the other keys. Nevertheless, that\nrequires that sensor nodes will be aware of their location.\n" }, { "page_number": 337, "text": "326\n■\nSecurity in Wireless Mesh Networks\nLDK [3] (Location Dependent Key management) uses random key pre-\ndistribution and does not require any knowledge about the deployment\nof the nodes. Nevertheless, it is only designed for static nodes and the\nauthor admits that is vulnerable during an interval after nodes deployment.\nIn addition, it assumes that certain special nodes that are also deployed\nrandomly (called anchors) are tamper proof and that each sensor node is\nin a transmission range of at least one anchor node.\nThe next section discusses the convenience of using asymmetric cryp-\ntography mechanisms instead of symmetric cryptography ones and the use\nof solutions that require tampering resistant nodes and misbehavior detec-\ntion schemes. Related work that is not strictly about key management, but\nabout securing the routing protocol, is discussed in Chapter 9.\n10.3\nPlaying without a Referee\n10.3.1\nSymmetric versus Asymmetric Cryptography\nIf in a wireless network all routing messages are encrypted with a symmetric\ncryptosystem, it means that everybody that we want to be able to participate\nin the network has to know the key. That is not a big problem if nodes\nare a “team” that gets to know the “team-key” before they are deployed or\ntry to interconnect, creating an ad hoc wireless network. A member of the\nteam trusts the other members of the team, so they assume that the other\nmembers of the team will not act in a malicious or selfish way. They trust\nthe other members and authorize them to change their routing tables.\nMaybe this is the best thing to do for military scenarios (besides the\nproblem of the compromised nodes and some others), but it is probably\nnot a good approach for a wireless network where everybody can partic-\nipate (like in a convention, in a meeting room, on a campus, or in our\nneighborhood). In this case there is a problem: nodes do not trust each\nother (and they should not). They are not a team. So what can be done?\nHow can everybody be forced to be honest? A possible approach is to only\nbelieve a piece of routing information if the originator of such information\nis the destination of the route. In this way, if a node lies, the only thing it\nwill achieve is that the other nodes will not be able to communicate with\nit (because you can only lie about yourself ).\nIn this kind of scenario, the best option is to use an asymmetric cryp-\ntosystem (with public and private key pairs) so that the originator of the\nroute messages signs its messages. It would not be needed to encrypt the\nrouting messages because routing messages are not meant to be secret.\nThe only requirement is that the nodes will be able to detect forged rout-\ning messages.\n" }, { "page_number": 338, "text": "Key Management in Wireless Mesh Networks\n■\n327\n10.3.2\nObscurity- and Tamper-Resistant Devices\nBecause there had not been a clear way to secure ad hoc networks, by the\nend of the last century, some people decided to dust off the tamper-resistant\napproaches. There are several papers [1,2,5] which discuss why “trusting\ntamper resistance is problematic.” The attacks against the supposededly\ntamper-resistant devices range from playing with things like voltage, tem-\nperature, fast signals, and clock frequency to affect EEPROM operation to\nthe use of chemicals to remove the covering plastic or the processors.\nThose papers show that obscurity is not the way to obtain security. They\nshow that there is no such thing as a tamper-resistant device. Therefore,\ntrying to combine symmetric cryptography solutions with tamper-resistant\ndevices to create the same result provided by alternatives that use asym-\nmetric cryptography does not make sense.\nIn addition, having a secret key stored in so many devices and with the\nproblem that, once the key is known to a malicious entity, the whole secu-\nrity of the network (not only the security of a single node) is compromised,\nmakes the whole approach too risky to be even seriously considered.\n10.3.3\nMisbehaving Detection Schemes\nIn the year 2000, a long trail of papers about how to secure ad hoc networks\nby using misbehavior detection schemes started (e.g., [17]). This kind of\napproach has two main problems:\n1.\nIt is quite likely that it will be not feasible to detect several kinds\nof misbehavior (especially because it is very hard to distinguish\nmisbehavior from transmission failures and other kind of failures).\n2.\nIt has no real means to guarantee the integrity and authentication\nof the routing messages.\nTherefore, unless those problems are addressed, this approach will not\nbe feasible. Any malicious node can generate forged misbehaving reports,\nmaking everybody believe that the rest of the nodes are even more evil\nthan itself. Trying to use reputation schemes is just a way of blurring the\nproblem.\n10.4\nThe Concept of Identity\nThe concept of identity in computer applications is most of the time binded\nto a person and, on occasion, to a program or to a process. But, in rout-\ning protocols it must be binded to the node itself as user and application\nidentification only makes sense at the application level.\n" }, { "page_number": 339, "text": "328\n■\nSecurity in Wireless Mesh Networks\n10.4.1\nIdentity in a Place without Authorities\nOne of the most important consequences of the nature of wireless networks\nis that one cannot assume that a node that is part of a network will be always\nreachable by all the other nodes. This implies that there cannot be central\nservers in the conventional meaning of fixed networks. Therefore, the use\nof CAs for wireless networks is not feasible.\nThe approach of distributing the CA functionality among ad hoc nodes\n(by dividing the private keys into shares) discussed in [28] implies a huge\noverhead, and it may be ineffective in a network where partitions occur or\nwhere there is high mobility. In addition, it will not work at all in trivial\nscenarios, like when a network partition is composed of only two or three\nnodes.\nThe use of key management protocols that require exchange of mes-\nsages between two nodes that need to forward routing information and\nthat might never see each other again is, most of the time, not a choice.\nIt would be great if the key management scheme would not need to send\nany additional messages besides the ones used for the routing protocol. Is\nall this possible?\n10.4.2\nMAC Addresses Are Not Unique Identifiers\nJust in case somebody does not know it yet, MAC addresses are not unique\nidentifiers. Moreover, you can change the MAC address (if you have the\nproper rights) of your network card under virtually any operating system.\nFor instance, in most Linux distributions you can just type this as root:\n/etc/init.d/networking stop\nifconfig eth0 hw ether 01:23:45:67:89:A0\n/etc/init.d/networking start\nIf you use Free BSD, you would type:\nifconfig fxp0 ether 01:23:45:67:89:A0\nAnd, if you use Mac OS X, you would type:\nsudo ifconfig en0 lladdr 01:23:45:67:89:A0\nYou can also change the MAC address under Windows®, although the\nmethod will vary depending on the version you use, and it is not going to\nbe as straightforward as in the UNIX world.\n" }, { "page_number": 340, "text": "Key Management in Wireless Mesh Networks\n■\n329\n10.4.3\nWhat Identifies Me?\nAnother characteristic of servers in fixed networks, besides continuous\navailability, is the fact that clients have to know the server’s IP address\n(or to know its human address and have the IP address of a DNS server).\nThe same thing happens in wireless networks for any node you want to\nmake a request to or initiate an exchange of data.\nHowever, current trends about addressing in ad hoc networks are driv-\ning toward dynamic address allocation and auto-configuration [6,25]. In\nthese schemes, typically a node picks a tentative address and checks if it is\nalready in use by broadcasting a query. If no conflict is found, the node is\nallowed to use that address. If a conflict is found, the node is required to\npick another tentative address and repeat the process.\nBut then, if IP addresses do not identify a node (because they are dy-\nnamically allocated), how does a node know the IP address of the node to\nwhich it wants to sent data? In fixed networks, if a node wants to send data\nto another one, it needs to know its address (it cannot send anything to a\nnode that has a dynamic address because it does not know its IP address).\nThe binding between public keys and other attributes is typically\nachieved by using public key certificates. In some limited scenarios, a pos-\nsible approach could be for a certification authority (that would live in a\nfixed network) to issue such certificates that the nodes could collect be-\nfore going to the wireless “playground.” However, this is not feasible for\na large group of the targeted scenarios. An added problem is that the IP\naddress should be one of the attributes binded to the public keys because\nit is binded to your identity.\nIn WMNs that are created in an ad hoc manner, node identity must\nbe its private key that can be used to sign messages and be verified by\nothers with the node’s public key. We say it must be their key pair because\nthere is nothing else. Another important observation is that, because we are\nworking at the routing layer, those key pairs identify not users, but nodes.\nThe problem with establishing public pairs as the identity of the nodes\nis the fact that one can generate as many key pairs as it desires. This,\ncombined with the fact that one can set its own MAC and IP addresses\nto the values it wants, can lead to a scenario where a malicious node\nhas different sets of key pairs, IP address, and MAC address to use as\ndifferent personalities. There is no easy way to detect that. But it is feasible\nto design a key management scheme that prevents one malicious node\nfrom impersonating another.\nTo sum up, what is required is a system that achieves the following: IP\naddresses will be assigned dynamically, nodes will be identifiable by their\nIP addresses, and a binding between the public key and the IP address\nof a node. All this should be achieved without any kind of certification\nauthorities, which is quite a challenge.\n" }, { "page_number": 341, "text": "330\n■\nSecurity in Wireless Mesh Networks\n10.5\nDynamically Generated IP Addresses\nThe proposal of SAKM is to generate IP addresses in a similar way [19]. In\nthat paper, they were using what they called SUCV (Statistically Unique and\nCryptographically Verifiable) addresses. SUCV addresses were designed to\nprotect binding updates in mobile IPv6. SUCV addresses are generated by\nhashing an “imprint” and the public key. That imprint (that can be a random\nvalue) is used to limit certain attacks related to mobile IP.\nFor wireless networks, it is only needed to hash the public key. The hash\ndigest (or a sub-string of it) may be formatted in some specific way (to be\na valid IP address), and will be a Cryptographically Generated Address\n(CGA), which will also be statistically unique. When a message that uses\nthe CGA as the source IP address and the public key of a node is signed\nby its private key, it can be verified by any other node that the node has a\ncertain identity (represented by the knowledge of the secret key).\n10.5.1\nSAKM IP Address Generation\nIn SAKM, it is recommended to use IPv6 (instead of IPv4) due to its bigger\naddress length (that would guarantee the statistical uniqueness of the IP\naddresses). The address can be, then, a network prefix of 64 bits with a 64-\nbit SAKM HID (Half IDentifier) or a 128-bit SAKM FID (Identifier). These\ntwo identifiers are generated almost in the same way as the sucvHID and\nthe sucvID in SUCV (with the difference that they hash the public key\ninstead of an imprint):\nS AK M H I D = S H A1H M AC 64(PublicK ey, PublicK ey)\nS AK M F I D = S H A1H M AC 128(PublicK ey, PublicK ey)\nThere will be a flag in the SAODV (or whatever other protocol that uses\nSAKM) routing message extensions (the H flag) that will be set to 1 if the\nIP address is an HID and to 0 if it is an FID.\nFinally, if it has to be a real IPv6 address, a couple of things should be\ndone [11]:\n■\nIf HID is used, then the HID behaves as an interface identifier and,\ntherefore, its sixth bit (the universal/local bit) should be set to zero\nto indicate local scope (because the IP address is not guaranteed to\nbe globally unique).\n■\nAnd, if FID is used, then a format prefix corresponding to the wire-\nless network should be overwritten to the FID. Format prefixes 010\nthrough 110 are unassigned and would take only three bits of the\nFID. Format prefixes 1110 through 1111 1110 0 are also unassigned\n" }, { "page_number": 342, "text": "Key Management in Wireless Mesh Networks\n■\n331\nand they would take between four and nine bits of the FID. All\nof these format prefixes need to have 64-bit interface identifiers in\nEUI-64 format, so universal/local bit should be set to zero.\nThe length of an IPv4 address is probably too short to provide the\nstatistical uniqueness that this scheme requires when the number of nodes\nis very big. Nevertheless, if the number of nodes is assumed to be low\nenough (around 100 nodes or less), it is not very unrealistic to expect that\nthe statistical uniqueness property will hold.\nThe SAKM IPv4 address will have a network prefix of eight bits and\nan SAKM 4ID (IPv4 Identifier). The network prefix can be any number\nbetween 1 and 126 (both included) with the exception of 14, 24, and 39\n[14]. The network prefix 10 can only be used if it is granted that it will not\nbe connected to any other network [23].\nThe SAKM 4ID will be the first bits of the SAKM HID and the H flag\nwill be set.\n10.5.2\nSAKM Message Fields\nThe public key should be included in the routing messages that are signed,\nso that the nodes can verify the signature. Because, obviously, the public\nkey should be signed by the signature, it is placed before the signature\nfield.\nThe identifier of the algorithm that is used to sign the message is\nspecified in the Signature Method field. The possible values are shown in\nTable 10.1 (being mandatory to support RSA). Because SAODV (or what-\never other protocol uses SAKM) could allow more than one possible sig-\nnature method, it might happen that a node has to verify a signature with\na method it does not know. If this happens, the node will consider that the\nverification of the signature has failed.\nThis implies that all the nodes that form part of a wireless network\nshould know all the methods used by all the other nodes to sign their\nTable 10.1\nPossible Values of the Signature\nMethod Field\nValue\nSignature Method\n0\nReserved\n1\nRSA [24]\n2\nDSA [26]\n3\nElliptic curve [15]\n4–127\nReserved\n128–255\nImplementation dependent\n" }, { "page_number": 343, "text": "332\n■\nSecurity in Wireless Mesh Networks\nTable 10.2\nPossible Values of the Hash F Sign Field\nHash F Sign\nHash Length\nValue\nRESERVED\n—\n0\nMD2\n(128 bits)\n1\nMD5\n(128 bits)\n2\nSHA1\n(160 bits)\n3\nSHA256\n(256 bits)\n4\nSHA384\n(384 bits)\n5\nSHA512\n(512 bits)\n6\nReserved\n7–127\nImplementation dependent\n—\n128–255\nmessages. This is not a problem because, typically, all nodes of a wireless\nnetwork will use the same method (or two different methods the most).\nThe fact that there is more than one possible signature method is because\ndifferent networks may have tighter security requirements than some others\nand, therefore, use different signature methods.\nThe same happens with the hash function used to generate the hash\nthat will be signed. The identifier of the hash algorithm is specified in\nthe Hash F Sign field. The possible values are shown in Table 10.2 (being\nmandatory to support SHA1).\nThe exact codification of the all the fields is shown in Section 10.8.\n10.6\nDuplicated Address Detection\nIf a node A receives a routing message that is signed by a node B that has\nthe same IP address as one of the nodes for which A has a route entry\n(node C), it will not process that routing message normally. Instead, it will\ninform B that it is using a duplicated IP and it will prove it by adding the\npublic key of C (so B can verify the truthfulness of the claim).\nWhen the node B receives a routing message that indicates that some-\nbody else has the same IP address as itself (or it realizes it by itself ), it\nwill have to generate a new pair of public/private keys. After that, it will\nderive its IP address from its public key and it might inform all the other\nnodes (through a broadcast) of its new IP address with a special message\nthat contains the two IP addresses (the old and the new ones) and the two\npublic signatures (old and new) signed with the old private key and the\nnew private key. Nevertheless, it is much better if that message is unicast\n(instead of broadcast) to all the nodes it considers should receive this in-\nformation (in the case they are just a few). This unicast will be answered\n" }, { "page_number": 344, "text": "Key Management in Wireless Mesh Networks\n■\n333\nwith an acknowledge message by the receiver if it verifies that everything\nis in order.\nAfter this, the node will generate a route error message for the old IP\naddress. Its propagation will delete the route entries for the old IP address\nand, therefore, eliminate the duplicated addresses. This route error message\nmay have a message extension that tells which is the new address. In this\nway, the nodes that receive the routing message can already create the\nroute to the new IP address.\nThis solution allows two nodes to coexist in the same network with the\nsame IP address until one of them realizes it. This can be considered as\na good trade-off between the impact of changing address (and having a\ncoexisting period of two nodes with the same IP address) and the extremely\nlow probability of having address collision.\nIntermediate nodes could decide to store the IP addresses and public\nkeys of all the nodes they would meet (or of the last N nodes, depending\non their capabilities); that would allow an earlier detection of duplicated\nIP addresses in the network.\nAn alternative to this solution could be that, when a node detects that\nanother node is using the same IP address, it would keep its public/private\nkey pair and change the used IP address by applying a salt to the algorithm\nthat derives the IP address from the public key. Salt variations of hash\nalgorithms have been used to avoid dictionary attacks of passwords [18].\nThe “salt” is a random string that is added to the password before being\nhashed. This idea can be adapted with a very different purpose. If the\nstatistically unique IP address is derived from the public key and a salt\n(instead of only from the public key), the node that detects or is informed\nthat its IP address is also used by another node can change its IP address\nwithout changing its public key by just changing the salt.\nNevertheless, that would imply that the salt used by a node should be\nincluded in all the routing messages and stored in all the entries of the\nrouting tables; and still, the node has to inform the others of its change of\nIP address. Therefore, it will not be used for the purpose of SAKM.\nIn conclusion, the approach described here handles properly the very\nunlikely situation of two nodes with the same IP address, without adding\nany complexity to the typical situation.\nThe format of the SAKM duplicate address detection messages is shown\nin Section 10.9.\n10.6.1\nDuplicated IP Address Detection for SAKM\nSAKM can deal with the duplicated IP address problem as described earlier.\nDuplicate address (DADD) detected message is sent to notify to a node that\nits address is already being used by another node. New address (NADD)\n" }, { "page_number": 345, "text": "334\n■\nSecurity in Wireless Mesh Networks\nnotification message is used to inform that the node has changed key pair\nand IP address. Finally, new address acknowledgment (NADD-ACK) mes-\nsage is used to confirm the reception of the NADD. In SAKM, NADD is\nalways unicast (never broadcast).\n10.6.2\nNetwork Leaders\nThe original SAODV design established that besides how key distribution\nis achieved, when distributing a public key, this should be binded to the\nidentity of the node (of course) and also to its netmask (in the case the node\nis a network leader). This was to prevent an attack in which a malicious\nnode becomes a black hole for a whole subnet by claiming that it is their\nnetwork leader.\nIn the new approach presented here, ad hoc nodes will typically never\nbe network leaders. Network leaders will be only fixed nodes that typically\ngive access to the fixed network and the nodes in the wireless network\nshould know their IP addresses, prefix size, and public keys.\nNetwork leaders will not change their IP address in case there is a\nnode that happens to generate the same IP address. A node generating its\nIP address will check if the resulting IP address corresponds to the network\nleader or to the subnet corresponding to its prefix size. A node detecting\nanother node using the network leader IP address or any of the ones corres-\nponding to the leader subnet will inform the node and not the network\nleader.\n10.7\nDelayed Verification of Signatures\nAs stated in the Introduction, there has been some concern (e.g., [12,13,21])\nthat using signatures might require a processing power that might be ex-\ncessive for certain kinds of ad hoc scenarios. Delayed verification addresses\nthis problem by revising one of SAODV’s security requirements from the\nlist that was stated in [9].\n10.7.1\nRevised Security Requirements\nThe security requirements that will be provided are source authentication\nand integrity (that combined provide data authentication) and delayed\nimport authorization. Import authorization was defined in [9] as the ultimate\nauthority about routing messages regarding a certain destination node be-\ning that node itself. Therefore, a node will only authorize route information\nin its routing table if that route information concerns the node that is send-\ning the information. In this way, if a malicious node lies about it, the only\n" }, { "page_number": 346, "text": "Key Management in Wireless Mesh Networks\n■\n335\nthing it will cause is that others will not be able to route packets to the\nmalicious node.\nDelayed import authorization allows route entries and route entry dele-\ntions in the routing table that are pending verification. They will be veri-\nfied whenever the node has spared processor time or before these entries\nshould be used to forward data packages.\nThe security requirements will not include confidentiality and non-\nrepudiation because they are not necessarily critical services in the con-\ntext of routing [10]. They will not include either availability (because an\nattacker can focus on the physical layer without bothering to study the\nrouting protocol) and they will not address the problem of compromised\nnodes (because it is arguably not critical in non-military scenarios).\n10.7.2\nAchieving Delayed Import Authorization\nIn reactive ad hoc routing protocols, most of the routing messages that\ncirculate in the network are (by far) route requests. This is due to the fact\nthat route requests are broadcast. Route replies are unicast back through\nthe selected path. Route error messages are unicast down through the tree\nof nodes that had a route to the now-unreachable node that is advertised\nby the route error message.\nWhen a node receives a routing message, it creates a new entry in its\nrouting table (the so-called reverse route). Therefore, after the broadcast of\nthe route request, all the nodes in the network (or in the broadcast ring)\nhave created reverse routes to the originator of the route request. From all\nthese reverse routes, most of them will expire soon (typically all but the\nones that are in the selected path through which the route reply will travel).\nThen, the question is why all these route requests should be verified\n(with the consequent delay in the propagation of the broadcast) when most\nof them are going to be soon discarded. The answer is that there is no need\nto verify them until the corresponding route reply comes back and the node\nknows that it is in the selected path. The other reverse routes will expire\nwithout being verified.\nActually, the two signatures (the ones from the route request and route\nreply) will be verified after the node has forwarded the route reply. In this\nway transmissions of the route requests and replies occur without any kind\nof delay due to the verification of the signatures.\nFollowing the same idea, the signature of route error messages (and in\ngeneral, any routing message that has to be forwarded) can also be verified\nafter forwarding them.\nRoutes pending verification will not be used to forward any packet. If\na packet arrives for a node for which there is a route pending verification,\nthe node will have to verify it before using that route. If the verification\nfails, it will delete the route and request a new one.\n" }, { "page_number": 347, "text": "336\n■\nSecurity in Wireless Mesh Networks\nUser space \nKernel space \nKernel routing table \n• AODV validated routes\n• Other protocol routes\n• AODV validated routes\nAODV routing table\nSAODV daemon\n• AODV non-validated routes\nFigure 10.1\nSAODV daemon.\n10.7.3\nSAKM with Delayed Verification\nWhen a node needs to send or to forward a packet to a destination for which\nit does not have an active route, first it will check if it has a route pending\nvalidation. If it does, it will try to validate it and, if it was successfully\nvalidated, it will mark it as active and use it. If after all this there is not an\nactive route, the node will start a route discovery process.\nAs shown in Figure 10.1, only once the validation is done successfully,\nthe route is incorporated in the routing table of the node. That avoids doing\ndirty hacks into the routing table of the operating system of the node. The\npackets can be routed normally, and only when there is a route lookup\nthat the routing table cannot resolve, the petition is captured by the SAODV\nrouting daemon.\nFigure 10.2 shows that in the case where there is a routing middleware\n(like Zebra1 or Quagga2), the middleware routing table will contain the\nvalidated routes from the SAODV daemon combined with the ones from\nthe other routing daemons, and the routing table in the kernel the ones\nwith lowest “administrative distance” (in case there is a route to the same\ndestination provided by two different routing daemons).\nTalking about administrative distances, none of the routing protocols for\nwireless networks that are being designed or standardized have specified\nwhich would be the appropriate administrative distance for them. Let us\nlook to the “standard de facto” (Cisco, Zebra, etc.) default administrative dis-\ntance values. Probably a good default distance value would be between 160\n1 www.zebra.org\n2 www.quagga.net\n" }, { "page_number": 348, "text": "Key Management in Wireless Mesh Networks\n■\n337\nKernel routing table\nRouting table\n• Routes selected by the lowest\n administration distance value\n• SAODV validated routes\n• Validated routes\n• Non-validated routes\nSAODV daemon\nOther routing\ndaemon\nRouting table\nRouting table\n• Other daemons, routes\nUser space\nKernel space\nRouting middleware\nFigure 10.2\nSAODV daemon with a routing middleware.\n(Cisco’s On-Demand Routing) and 170 (external routes in EIGRP). There-\nfore, a default distance value of 165 for SAODV (and also for AODV in\ngeneral) would be appropriate.\n10.8\nSAKM Encoding of Public Keys and Signatures\nThis section is provided for completeness, and it shows how public keys\nand signatures are encoded under SAKM. When SAODV is used in conjunc-\ntion with SAKM, it will encode the originator public key for each routing\nmessage before its signature field.\nFigure 10.3 and Table 10.3 show the fields of the encoding of the sig-\nnature. Figure 10.4 and Tables 10.4 and Table 10.5 show the fields of the\nencoding of the public key.\n10.9\nSAKM Duplicate Address Detection Messages\nThis section serves as a reference of the SAKM duplicate address detection\nmessages structure. It shows their fields and what they are used for.\nFigure 10.5 and Table 10.6 show the fields of the duplicated address\n(DADD) detected message. Figure 10.6 and Table 10.7 show the fields of\n" }, { "page_number": 349, "text": "338\n■\nSecurity in Wireless Mesh Networks\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\nLength\nPadd length\nReserved\nReserved\nPublic key\nPadding (optional)\nHash F Sign\nSign method\nH\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n. . .\n. . .\nFigure 10.3\nEncoding of the signature.\nthe new address (NADD) notification message. And, finally, Figure 10.7 and\nTable 10.8 show the fields of the new address acknowledgment (NADD-\nACK) message.\nTable 10.3\nThe Fields of the Encoding of the Signature\nField\nValue\nSignature method\nThe signature method used to compute the signatures. (RSA is\nencoded as 1)\nH\nHalf Identifier flag. Set to 1 indicates the use of HID; set to 0,\nthe use of FID\nReserved\nSent as 0; ignored on reception\nPadding length\nSpecifies the length of the padding field in 32-bit units. If the\npadding length field is set to zero, there will be no padding\nHash F Sign\nThe hash function used to compute the hash that will be signed.\nBecause, typically you do not want to sign the whole message,\nyou sign a hash of the message. (MD5 is encoded as 2 and\nSHA1 is encoded as 3)\nReserved\nSent as 0; ignored on reception\nLength\nThe length of the Value field (not including the Length and\nReserved fields) in 32-bit units\nPublic key\nThe public key of the originator of the message. This field has\nvariable length, but it must be 32-bits aligned\nPadding\nRandom padding. The size of this field is set in the Padding\nLength field\n" }, { "page_number": 350, "text": "Key Management in Wireless Mesh Networks\n■\n339\nLength\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\nReserved\nModulus\nExp\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\nFigure 10.4\nEncoding of the public key.\nTable 10.4\nThe Encoding of an RSA Public Key\nField\nValue\nReserved\nSent as 0; ignored on reception\nLength\nThe length of the Modulus field (not including the Length and Reserved\nfields) in 32-bit units\nExp\nThe Exponent (e) encoded as specified in the next table\nTable 10.5\nThe Encoding of the RSA Exponent\n00\nThe components are encoded in the standard way. The Exponent (e) will be\nspecified after the Modulus (n)\n01\nSpecifies that Exponent (e) is 65537\n10\nSpecifies that Exponent (e) is 17\n11\nSpecifies that Exponent (e) is 3\nNote: A message that uses any of these “smartly chosen”exponents must include random\npadding (in the Padding field). There is no security problem with everybody using the\nsame exponent\nReserved\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\nType\nLength\nH\nDuplicated node’s IP address\nDuplicated node’s Public key\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\nFigure 10.5\nDuplicated address (DADD) detected message.\n" }, { "page_number": 351, "text": "340\n■\nSecurity in Wireless Mesh Networks\nTable 10.6\nDuplicated Address Detected Message Fields\nField\nValue\nType\n64\nLength\nThe length of the type-specific data, not including\nthe Type and Length fields of the message\nH\nHalf Identifier flag; set to 1 indicates the use of\nHID, set to 0, the use of FID\nReserved\nSent as 0; ignored on reception\nDuplicated node’s IP address\nThe IP address of the node that uses a duplicated\nIP address\nDuplicated node’s public key\nThe public key of the node that uses a duplicated\nIP address\n10.10\nSimulation Results\nThe purpose of using SAODV with delayed verification is to obtain the same\nlevel of security as with the original SAODV, but without its main draw-\nbacks. These drawbacks are a quite bigger average end-to-end delay and\na higher power consumption by the nodes (when compared with AODV).\nReserved\nReserved\nPadd length\nH\nSign method\nReserved\nPadding (optional)\nOld public key\nPadd length 2\n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\nType\nLength\nSignature with new key\nSignature with old key\nPadding 2 (optional)\nNew public key\nSign method 2\nH\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\nFigure 10.6\nNew address (NADD) notification message.\n" }, { "page_number": 352, "text": "Key Management in Wireless Mesh Networks\n■\n341\nTable 10.7\nNew Address Notification Message Fields\nField\nValue\nType\n65\nLength\nThe length of the type-specific data, not including the\nType and Length fields of the message\nReserved\nSent as 0; ignored on reception\nSignature method\n. . . padding\nThe same in the message extensions. Corresponds to the\nSignature with Old Public Key signature\nSignature method 2\n. . . padding 2\nThe whole block of fields is repeated. Corresponds to the\nSignature of the New Public Key signature\nSignature with old key\nThe signature (with the old key) of all the fields in the\nAODV packet that are before this field\nSignature with new key\nThe signature (with the new key) of all the fields in\nthe AODV packet that are before this field\nThese drawbacks are due to the computation of asymmetric cryptography\nprimitives (message signature and verification). Through the use of simu-\nlations, it was shown that delayed verification actually achieves this.\nThe simulations were done with 30 nodes moving at a maximum speed\nof 10 meters per second in a square of 1000 × 1000 meters. They simu-\nlated the establishment of ten connections that started between second 0\nReserved\nH\nSign method\nNew IP address\nOld IP address\nPadding (optional)\nPadd length \n5\n4\n3\n2\n1\n1\n0\n9\n8\n7\n6\n5\n4\n3\n2\n1\n0\n0\n6 7 8 9 0 1 2 3 4 5 6 7 8 9 0\n3\n1\n2\nType\nLength\nReserved\nSignature\nPublic key\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n. . .\n. . .\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\n+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+‒+ +\n‒\nFigure 10.7\nNew address acknowledgment (NADD-ACK) message.\n" }, { "page_number": 353, "text": "342\n■\nSecurity in Wireless Mesh Networks\nTable 10.8\nNew Address Acknowledgment Message Fields\nField\nValue\nType\n66\nLength\nThe length of the type-specific data, not including\nthe Type and Length fields of the message\nReserved\nSent as 0; ignored on reception\nOld IP address\nThe old IP address\nNew IP address\nThe new IP address\nSignature method\n. . . padding\nThe same in the message extensions\nSignature\nThe signature of all the fields in the AODV packet that are\nbefore this field\nand second 25 (according to an uniform distribution) and ended at the\nend of the simulation. The simulation time was of 100 seconds, and the\nconnections where constant bit rate (a packet of 512 each 0.25 seconds).\nThe nodes in the simulations have used as routing protocols: plain\nAODV, SAODV with RSA, SAODV with ECC (Elliptic Curve Cryptography),\nand SAODV with delayed verification (SAODV2 in the figure) with ECC.\nThere is no point in using delayed verification with RSA because its verifica-\ntion time is completely negligible (delayed verification reduces the amount\nof verifications that have to be done). That means that SAODV with RSA\nwith or without delay verification will give practically identical results. RSA,\nDSA, and ECC have been used with key lengths that provide equivalent\nsecurity (1368 bits for RSA and DSA, and 160 bits for ECC).\nTable 10.9 shows the times for signing/verifying in a Compaq iPAQ 3670\n(206 MHz, 16 M ROM, 64 M RAM) according to [27]. DSA is not used in the\nsimulations as it presents the worst of RSA and ECC (slow signature and\nverification, and fast increase of computational overhead as the key length\nneeds to be bigger).\nIn the simulations, end-to-end delay of the packets, packet delivery\nfraction, and normalized routing load were measured. Figure 10.8 shows\nTable 10.9\nTimes for a Compaq\niPAQ 3670\nRSA\nDSA\nECC\nKey length\n1368\n1368\n160\nSign\n210\n90\n42\nVerify\n6\n110\n160\n" }, { "page_number": 354, "text": "Key Management in Wireless Mesh Networks\n■\n343\n0.450 \n0.400 \n0.350 \n0.300 \n0.250 \n0.200 \n0.150 \n0.100 \n0.050 \n0.000 \nAODV \nSAODV \nRSA \nSAODV \nECC \nSAODV2 \nECC \nFigure 10.8\nSimulation results. Average end-to-end delay, measured in milli-\nseconds.\nthe averaged result of the end-to-end delay in data packet transmission.\nThere were practically no differences among the routing protocols in packet\ndelivery fraction (that was around 90 percent) and in normalized routing\nload (that was around 1).\nOne could expect quite different results with some other simulation\nscenarios, but almost always having SAODV with delayed verification and\nECC as the best of the SAODV options and with a performance very close\nto plain AODV.\nOne could argue that, in scenarios in where the routes have more hops,\nthe results of SAODV with delayed verification will be quite worse. But,\nactually, the results do not depend that much on the number of hops. This\nis due to the fact that intermediate nodes forward the RREP before verifying\nthe signatures of the RREQ and RREP. Therefore, it is most probable that\nby the time the node that forwards the RREP to the final destination verifies\nthe signatures of the RREQ and RREP, all the nodes of the route will also\nhave verified them.\nIn the future, when longer keys are needed, ECC results will look even\nbetter than with the key lengths used in these simulations. This is due to\nthe fact that, as the key size increases, the computational overhead of ECC\nincreases in a much slower manner than for RSA.\nTherefore, these simulations have shown that SAODV used with de-\nlayed verification and ECC performs better than the other combinations\nwith SAODV and that the performance penalty it introduces is almost\nnegligible.\n" }, { "page_number": 355, "text": "344\n■\nSecurity in Wireless Mesh Networks\n10.11\nOpen Issues\nAlthough it is true that there is no way to preclude a node of inventing\nmany identities, that cannot be used to create an attack against the secure\nrouting algorithm. An attacker cannot supplant another node, and a node\ncan always prove that it is the same node.\nDelayed verification makes possible that a malicious node creates in-\nvalid route requests that could flood the network. But, the same malicious\nnode can flood the network with perfectly valid route requests, and there\nwould be no easy way to know if it is trying to flood the network or if it is\njust trying to see if any of its friend nodes are present in the network (for\ninstance).\nAs explained before, an attacker cannot forge a public/private key pair\nfrom an IP address, so the identity token becomes the IP address itself.\nUsers of nodes might have a mechanism outside the network to bind their\npublic key to their physical identity.\nWith the current technology, SAODV with delayed verification and ECC\nprovides security features to AODV with an almost negligible performance\npenalty.\nIn the future, when longer keys are required, the gain of using delayed\nverification in conjunction to ECC compared to other SAODV options will\nbe even bigger than it is now. This is due to the fact that as key length gets\nbigger, the cost of signing/verifying in RSA and other cryptoalgorithms in-\ncreases exponentially as in ECC (for the equivalent key length): it increases\nin a logarithmic way.\nReferences\n[1]\nR. Anderson and M. Kuhn. Tamper resistance—a cautionary note. Proceed-\nings of the Second Usenix Workshop on Electronic Commerce, November\n1996.\n[2]\nR. Anderson and M. Kuhn. Low cost attacks on tamper resistant devices.\nIn IWSP: International Workshop on Security Protocols, LNCS, 1997.\n[3]\nF. Anjum.\nLocation dependent key management using random key-\npredistribution in sensor networks. In WiSe ’06: Proceedings of the 5th ACM\nWorkshop on Wireless Security, pp. 21–30, New York, 2006, ACM Press.\n[4]\nS. Basagni, K. Herrin, D. Bruschi, and E. Rosti.\nSecure pebblenets.\nIn\nProceedings of the 2001 ACM Iternational Symposium on Mobile Ad Hoc\nNetworking & Computing, MobiHoc 2001, pp. 156–163, Long Beach, CA,\nOctober 4–5, 2001.\n[5]\nE. Biham and A. Shamir. Differential fault analysis of secret key cryptosys-\ntems. In CRYPTO, pp. 513–525, 1997.\n[6]\nS. Cheshire and B. Aboba. Dynamic configuration of Ipv4 link-local ad-\ndresses. IETF INTERNET DRAFT—Work in progress, Zeroconf Working\nGroup, June 2001, draft-ietf-zeroconf-ipv4-linklocal-03.txt.\n" }, { "page_number": 356, "text": "Key Management in Wireless Mesh Networks\n■\n345\n[7]\nW. Du, J. Deng, Y. S. Han, S. Chen, and P. K. Varshney. A key management\nscheme for wireless sensor networks using deployment knowledge.\nIn\nINFOCOM, 2004.\n[8]\nL. Eschenauer and V. Gligor. A key management scheme for distributed\nsensor networks, 2002.\n[9]\nM. Guerrero Zapata and N. Asokan. Securing ad hoc routing protocols. In\nProceedings of the 2002 ACM Workshop on Wireless Security (WiSe 2002),\npp. 1–10, September 2002.\n[10]\nR. Hauser, A. Przygienda, and G. Tsudik. Reducing the cost of security in\nlink state routing. In Symposium on Network and Distributed Systems Secu-\nrity (NDSS ’97), pp. 93–99, San Diego, California, February 1997, Internet\nSociety.\n[11]\nR. Hinden and S. Deering. IP Version 6 Addressing Architecture. RFC 2373,\nJuly 1998.\n[12]\nY. C. Hu, D. Johnson, and A. Perrig. SEAD: Secure efficient distance vec-\ntor routing for mobile wireless ad hoc networks. In 4th IEEE Workshop\non Mobile Computing Systems and Applications (WMCSA ’02), June 2002,\npp. 3–13, June 2002.\n[13]\nY. C. Hu, A. Perrig, and D. Johnson. Ariadne: A secure on-demand routing\nprotocol for ad hoc networks. Technical report TR01-383, Rice University,\nDecember 2001.\n[14]\nIANA. Special-use IPv4 Addresses. RFC 3330, September 2002.\n[15]\nR Laboratories. Elliptic Curve Cryptography Standard. Public-Key Cryptog-\nraphy Standards (PKCS) January 13, 1998.\n[16]\nD. Liu and P. Ning. Location-based pairwise key establishments for static\nsensor networks. In SASN ’03: Proceedings of the 1st ACM Workshop on\nSecurity of Ad hoc and Sensor Networks, pp. 72–82, 2003. ACM Press, New\nYork.\n[17]\nS. Marti, T. J. Giuli, K. Lai, and M. Baker. Mitigating routing misbehavior in\nmobile ad hoc networks. In Proceedings of the 6th Annual International\nConference on Mobile Computing and Networking, pp. 255–265, 2000.\n[18]\nA. J. Menezes, P. C. van Oorschot, and S. A. Vanstone. The Handbook of\nApplied Cryptography, CRC Press, Boca Raton, FL, 1996.\n[19]\nG. Montenegro and C. Castelluccia. Statistically unique and cryptographi-\ncally verifiable (SUCV) identifiers and addresses. Network and Distributed\nSystem Security Symposium (NDSS ’02), February 2002.\n[20]\nG. O’Shea and M. Roe. Child-proof authentication for MIPv6 (CAM). ACM\nComputer Communication Review, April 2001.\n[21]\nP. Papadimitratos and Z. J. Haas. Secure routing for mobile ad hoc net-\nworks. SCS Communication Networks and Distributed Systems Modeling\nand Simulation Conference (CNDS 2002), January 2002.\n[22]\nA. Perrig, R. Szewczyk, J. D. Tygar, V. Wen, and D. E. Culler. SPINS: Security\nprotocols for sensor networks. Wireless Networks, 8(5): 521–534, 2002.\n[23]\nY. Rekhter, B. Moskowitz, D. Karrenberg, G. J. de Groot, and E. Lear. Ad-\ndress allocation for private internets. RFC 1918, February 1996.\n[24]\nR. Rivest, A. Shamir, and L. Adleman. A method for obtaining digital signa-\ntures and public-key cryptosystems. Communications of the ACM, 21(2),\nFebruary 1978.\n" }, { "page_number": 357, "text": "346\n■\nSecurity in Wireless Mesh Networks\n[25]\nS. Thomson and T. Narten. IPv6 stateless address autoconfiguration. IETF,\nRFC 2462, December 1998.\n[26]\nU.S. National Institute of Standards and Technology, Computer Systems\nLaboratory. Digital Signature Standard (DSS). Federal Information Process-\ning Standards Publication (FIPS PUB) 186, May 1994.\n[27]\nJ. Walter, J. Oleksy, and J. Kong.\nThe role of ECDSA in wireless com-\nmunications. Masters thesis. Computer Science Department. University of\nCalifornia, 2002.\n[28]\nL. Zhou and Z. J. Haas. Securing ad hoc networks. IEEE Network Magazine,\n13(6): 24–30, November/December 1999.\n" }, { "page_number": 358, "text": "SECURITY\nSTANDARDS,\nAPPLICATIONS,\nAND ENABLING\nTECHNOLOGIES\nIII\n" }, { "page_number": 359, "text": "" }, { "page_number": 360, "text": "Chapter 11\nSecurity in Wireless PAN\nMesh Networks\nStefaan Seys, Dave Singel´ee, and Bart Preneel\nContents\n11.1\nIntroduction ........................................................ 350\n11.1.1\nBasic Principles of Bluetooth ............................. 351\n11.1.2\nBasic Principles of ZigBee ................................ 352\n11.1.3\nDesigning a WPAN Security Architecture................. 353\n11.2\nBluetooth Security .................................................. 355\n11.2.1\nBluetooth Cryptographic Primitives ...................... 355\n11.2.2\nKey Agreement Protocol in Bluetooth.................... 358\n11.2.2.1 Generation of the Unit Key ..................... 358\n11.2.2.2 Generation of the Initialization Key ............ 358\n11.2.2.3 Mutual Entity Authentication .................... 359\n11.2.2.4 Generation of the Link Key ..................... 360\n11.2.2.5 Generation of the Encryption Key and\nthe Key Stream .................................. 361\n11.2.3\nSecurity Weaknesses in the Bluetooth Security\nArchitecture................................................ 361\n11.2.3.1 Unit Key ......................................... 362\n11.2.3.2 Location Privacy ................................. 362\n11.2.3.3 Security Depends on Security of PIN ........... 362\n11.2.3.4 Denial-of-Service Attacks........................ 363\n11.2.3.5 Encryption Algorithm E0 ........................ 364\n11.2.3.6 Bluejacking ...................................... 364\n349\n" }, { "page_number": 361, "text": "350\n■\nSecurity in Wireless Mesh Networks\n11.2.3.7 Implementation Errors .......................... 365\n11.2.3.8 Other Security Problems ........................ 365\n11.2.4\nBluetooth Security in Practice............................. 366\n11.3\nZigBee Security ..................................................... 366\n11.3.1\nZigBee Cryptographic Primitives ......................... 368\n11.3.1.1 CCM ∗Algorithm ................................. 368\n11.3.1.2 The AES Algorithm .............................. 369\n11.3.2\nSecurity Architecture of ZigBee ........................... 370\n11.3.2.1 Key Hierarchy ................................... 370\n11.3.2.2 ZigBee Trust Center ............................. 371\n11.3.3\nSecurity Weaknesses in the ZigBee Security\nArchitecture................................................ 372\n11.3.3.1 IV (Nonce) Management Problems ............. 372\n11.3.3.2 Improper Support of Group Keying ............ 373\n11.3.3.3 Key Management ................................ 373\n11.3.3.4 Replay Attacks ................................... 374\n11.3.3.5 Initialization Procedure.......................... 374\n11.3.3.6 Location Privacy ................................. 375\n11.3.3.7 Insufficient Integrity Protection ................. 375\n11.4\nConclusion and Open Issues ...................................... 376\nReferences................................................................. 376\nIn this chapter we analyze the security issues related to wireless personal\narea mesh networks. We start with a general introduction on wireless PAN\nnetworks and discuss the two most common technologies: Bluetooth and\nZigBee. Subsequently, we discuss the security architecture and the security\nweaknesses of Bluetooth and ZigBee, offering advice on how these weak-\nnesses could be mitigated. Finally, we conclude with challenging open\nresearch issues.\n11.1\nIntroduction\nAs more and more mobile devices (i.e., digital cameras, cell phones, GPS\nreceivers) became available on the market, it became apparent that en-\nabling these devices to communicate over wireless links would allow these\ndevices to work together and augment their functionality. In response to\nthis demand of a low-power wireless transmission medium, the Bluetooth\nSpecial Interest Group (SIG) was founded in 1998. Bluetooth is essentially\na cable-replacement technology that allows for a limited number of devices\nto communicate with each other via a wireless link.\nWith further miniaturization of electronic devices, it now becomes possi-\nble to manufacture tiny sensor and actuator nodes programmed to provide\n" }, { "page_number": 362, "text": "Security in Wireless PAN Mesh Networks\n■\n351\nspecific information (e.g., room temperature, light intensity, etc.) or per-\nform specific tasks (e.g., toggle lights, turn on sprinkler systems, etc.). If\nthese sensors can communicate using wireless links and automatically set\nup large ad hoc networks, this would drastically reduce the costs of deploy-\nment. Engineers and researchers soon discovered that Bluetooth or WiFi\nwould not be suitable for this task due to many reasons, the most impor-\ntant being power consumption and the lack of autonomous self-organized\noperation. This resulted in the IEEE 802.15.4 standard (among others) that\nwas completed in 2003. The ZigBee standard that specifies a set of higher\nlayer protocols to operate on top of IEEE 802.15.4 was released to the\npublic in 2005.\nThe main differences between Bluetooth and ZigBee are (1) ZigBee\nis more efficient and allows longer battery lifetime at the cost of lower\ntransmission speeds (100 to 1000 days for ZigBee compared to a couple of\ndays for Bluetooth); (2) Bluetooth only supports networks up to 8 nodes,\nwhile ZigBee supports up to 65,536 nodes; and (3) the range of ZigBee\n(30 m) is larger than the range of Bluetooth (10 m). These differences\nshow that ZigBee is targeted at large control and monitoring networks that\nshould be able to operate for years without maintenance, while Bluetooth\nis a cable replacement technology that is used between devices that can be\nregularly recharged.\nIt is clear that providing security for both types of networks is essen-\ntial as wireless links are easy to eavesdrop undetected. The fact that these\nnetworks run on battery-operated devices with limited processing power\nmeans that the security solutions should be as efficient as possible and\navoid intensive use of expensive cryptographic operations such as public\nkey encryption or digital signatures. Moreover, these networks normally\noperate autonomously without access to online key servers or certification\nauthorities. This means that conventional means of key establishment are\nnot always applicable to these networks. To make things even more dif-\nficult, ZigBee networks allow multi-hop routing and node mobility. This\nmeans that nodes do not have a clear idea of the continuously changing\nnetwork topology. These specific properties present interesting challenges\nwhen designing security and privacy solutions in these environments. In\nthis chapter, we investigate how Bluetooth and ZigBee have implemented\ntheir security architecture.\n11.1.1\nBasic Principles of Bluetooth\nIn February 1998, the Bluetooth SIG [1] was founded by major players\nin the telecommunications and network industries: Ericsson, IBM, Intel,\nNokia, and Toshiba. In the next six years, several other companies joined\nthe SIG and now there are already more than 3000 members. The major\ntask of this organization was the creation of the Bluetooth specification\n" }, { "page_number": 363, "text": "352\n■\nSecurity in Wireless Mesh Networks\nwhich describes how mobile phones, computers, PDAs, headsets, and other\nmobile devices can communicate with each other over a wireless link. In\n2000, the Bluetooth standard was included in IEEE 802.15 [2], the Wireless\nPersonal Area Network (WPAN) Working Group. The specifications have\nbeen updated several times: the latest version is v2.0, which was published\nin 2004.\nThe Bluetooth wireless technology [3,4] realizes a low-cost, short-range\nwireless voice- and data-connection through radio propagation. The pri-\nmary use of Bluetooth is cable replacement, most suited for small networks\nwith relatively high load of communication over short distances. With a\nnormal antenna, the maximal range is about 10 m. The Bluetooth wire-\nless technology uses the 2.4 GHz band, which is unlicensed, and can be\nused by many other types of devices such as cordless phones, microwave\novens, WiFi [5], and baby monitors. Any device designed for use in an un-\nlicensed band should provide robustness in the presence of interference,\nand the Bluetooth wireless technology has many features to achieve this,\nincluding spread spectrum and frequency hopping. Every time a Bluetooth\nwireless link is formed, it is within the context of a piconet. A piconet\nconsists of maximally eight devices that occupy the same physical channel.\nIn each piconet, there is exactly one master, the other devices are called\nslaves. The theoretical maximum bandwidth is 1 Mbps. The real bandwidth\nis lower because of error correction. One of the main differences between\nBluetooth and some other wireless technologies is the ability to connect\ndifferent types of devices (e.g., a mobile phone with a PDA).\nIt is possible to configure the “visibility” of a Bluetooth device. When\na device is in non-discoverable mode, it does not respond to inquiries of\nother devices. When the device is in limited discoverable mode, it is dis-\ncoverable only for a limited period of time, during temporary conditions or\nfor a specific event. And finally, when it is in general discoverable mode,\nit is discoverable (visible) continuously. Each device is characterized by a\nfactory-established 48-bit identifier, unique for every device: the Bluetooth\nhardware address.\n11.1.2\nBasic Principles of ZigBee\nZigBee [6] is a specification set of high-level communication protocols that\noperate on top of the low-power Media Access Control (MAC) and Physi-\ncal (PHY) layers described in the IEEE 802.15.4 standard for WPANs [2]. In\n2003, the IEEE 802.15.4-2003 standard [7] was approved by the TG4 Task\nGroup of the IEEE 802.15 Working Group. The ZigBee v1.0 specifications\nwere ratified in 2004, based on the IEEE 802.15.4-2003 standard. The TG4\nTask Group put itself into hibernation in 2004, after forming the TG4b Task\nGroup. The task of TG4b is to write a revision for specific enhancements\nand clarifications of the IEEE 802.15.4-2003 standard. The ZigBee alliance\n" }, { "page_number": 364, "text": "Security in Wireless PAN Mesh Networks\n■\n353\nis now working on the v1.1 specifications that will benefit from these im-\nprovements proposed by the 802.15.4b Task Group.\nZigBee is aimed at extending battery lifetimes of low-power devices.\nThe primary use of ZigBee is control and monitoring in wireless sensor\nnetworks, most suited for large networks with small load of communication\nover short distances. The maximum range is about 30 m and the theoretical\nmaximum bandwidth is 250 kbps. ZigBee operates in the same unlicensed\n2.4 GHz radio band as Bluetooth. The radios use direct-sequence spread\nspectrum coding to avoid interference. The technology is intended to be\nsimpler and cheaper than other WPANs such as Bluetooth. The most capa-\nble ZigBee node type is said to require only about 10 percent of the software\nof a typical Bluetooth or Wireless Internet node, while the simplest nodes\nare about 2 percent. However, actual code sizes are much higher, more like\n50 percent of Bluetooth code size. ZigBee chip vendors have announced\n128-kilobyte devices. ZigBee uses two kinds of addressing: a 64-bit IEEE\naddress that can be compared to the IP address on the Internet and a 16-bit\nshort address. The short addresses are used once a network is set up. A\nnetwork can consist of maximally 216 = 65,536 devices.\nThere are three different types of ZigBee devices:\n1.\nZigBee coordinator : The most capable device, the coordinator,\nforms the root of the network tree and might bridge to other net-\nworks. There is exactly one ZigBee coordinator in each network. It\nis able to store information about the network, including acting as\nthe repository for keys. It configures the security level of the net-\nwork and the address of the trust center. Each network has exactly\none ZigBee trust center. This device is trusted by all other devices\nwithin the ZigBee network and is responsible for distributing and\nestablishing keys in the network. By default, the ZigBee coordinator\nis the ZigBee trust center. The coordinator can always designate an\nalternate trust center. Section 11.3.2 will focus more on the role of\nthe ZigBee trust center.\n2.\nA ZigBee router\ncan act as an intermediate router, passing data\nfrom other devices.\n3.\nA ZigBee end device contains just enough functionality to talk to\nits parent node (either the coordinator or a router). It cannot relay\ndata from other devices. It requires the least amount of memory,\nand therefore can be less expensive to manufacture than a ZigBee\nrouter or coordinator.\n11.1.3\nDesigning a WPAN Security Architecture\nA security architecture is a collection of building blocks and security policies\nthat make up a complete security solution. When designing a security\n" }, { "page_number": 365, "text": "354\n■\nSecurity in Wireless Mesh Networks\narchitecture for a specific technology (e.g., Bluetooth or ZigBee), one usu-\nally starts by performing a threat analysis. The resulting security require-\nments of this threat analysis are a set of inputs that are used to obtain the\ncomplete set of functional requirements for the security architecture. Other\ninputs include specific user requests, requirements to fulfill an existing API\nof existing applications, etc. Next to the functional requirements, the design\nof the security architecture has to take into account the specific properties\nand limitations of the platform it will run on.\nUsually a security architecture consists of a layered structure of differ-\nent building blocks, where a higher layer builds on the services offered by\nthe lower layer. The bottom layer consists of specific implementations (in\nhardware or software) of cryptographic algorithms such as the Bluetooth\nSAFER+ block cipher. One level up, we can make an abstraction of the\nspecifics of the cipher and supply abstract cryptographic primitives such\nas “block cipher” or “digital signature algorithm” to the higher layer. Using\nthese primitives, it is possible to implement cryptographic services such as\nauthentication, encryption, non-repudiation, etc. Going one layer higher,\nwe can use these cryptographic services to build more advanced security\nmechanisms such as end-to-end security (e.g., SSH or IPSec), electronic\npayment schemes, digital credentials, PKI, key management, etc. Note that\nthe latter two, PKI and key management, are built on top of cryptographic\nservices such as authentication and encryption, but are also required to\nexchange the keys that are used by the cryptographic services. Finally, one\ncan build complete applications on top of the provided security mecha-\nnisms. Obviously, the services we can offer at one layer are limited by\nthe services offered by the lower layer. Thus the choice of cryptographic\nalgorithms implemented on a certain platform reflects on the final secu-\nrity mechanisms that can be offered. For example, one cannot offer digital\ncredentials when no digital signature algorithm is available at the bottom\nlayer.\nThe major constraint when designing a security architecture for mobile\ndevices with limited resources (the target devices of both Bluetooth and\nZigBee) is available energy (i.e., battery power) and speed of the CPU.\nToday this prohibits the use of public key cryptography in the core of the\nsecurity architecture. Even elliptic curve-based algorithms are still orders\nof magnitude slower than, for example, the Advanced Encryption Standard\n(AES) [8,9]. A second important design factor for Bluetooth and 802.15.4\nis the fact that they are situated at the MAC layer of the OSI model. The\nMAC layer has limited functionality concerning communications, and the\nsecurity architecture should not out grow this functionality.\nFor the IEEE 802.15.4 standard the design is very clear. It provides the\nfour basic security services: message authentication, message integrity, mes-\nsage confidentiality, and replay protection. These services are all based on\n" }, { "page_number": 366, "text": "Security in Wireless PAN Mesh Networks\n■\n355\nthe AES block cipher. A higher layer can request four different security set-\ntings: no security, encryption only, authentication only, and encryption and\nauthentication (using AES-CCM). Obviously these services require crypto-\ngraphic keys to operate, but establishing these keys is not part of the IEEE\n802.15.4 security architecture and must be provided by the higher layers.\nBluetooth, also a MAC-layer system, does not provide the four basic se-\ncurity services, but does include a mechanism to bootstrap the system based\non a shared PIN-code (see Section 11.2.2). Bluetooth does not provide\nmessage authentication, meaning that an adversary could alter messages\nwithout detection or replay previous messages. However, it can protect\nthe confidentiality of messages. Next to this, there are also differences in\nthe implementation of the Bluetooth and IEEE 802.15.4 security algorithms.\nBluetooth uses the E 0 stream cipher (instead of the AES block cipher) for\ndata encryption. E 0 was designed to achieve a high energy efficiency with\na small hardware footprint, rather than for speed. Next to this stream ci-\npher, Bluetooth also uses the SAFER+ block cipher for key derivation (it is\ncommon practice to use block ciphers for key derivation). Normally E 0 is\nimplemented in hardware, while SAFER+ is implemented in software as it\nis only used when a new key needs to be negotiated.\nZigBee operates at higher layers (up to the application layer) on top\nof the IEEE 802.15.4 standard. The ZigBee security architecture provides\nnodes with a mechanism to establish keys with other nodes in the net-\nwork. Essentially, two different keys are known in ZigBee: a networkwide\nbroadcast key and link keys that allow two devices to set up end-to-end se-\ncurity (note that in practice there are more keys; see Section 11.3.2). These\nkeys are always established using a third party: the trust center of the net-\nwork (note that Bluetooth slaves establish keys with each other without\nthe use of the master in the piconet). Another important aspect of ZigBee\nsecurity is that every layer originating a frame is responsible for securing\nit. This simplifies the system, because multiple layers are not responsible\nfor securing the same frame. Next to this, all layers are allowed to use the\nsame key that is shared between source and destination (open trust model).\nFinally, ZigBee limits the encryption mode of IEEE 802.15.4 to CCM∗(see\nSection 11.3.1).\n11.2\nBluetooth Security\n11.2.1\nBluetooth Cryptographic Primitives\nBluetooth uses the synchronous stream cipher [10] E 0 to encrypt data\npackets. This encryption engine of Bluetooth is schematically depicted\nin Figure 11.1 [11,12]. E 0 is an autonomous Finite State Machine (FSM).\n" }, { "page_number": 367, "text": "356\n■\nSecurity in Wireless Mesh Networks\nLFSR 2\nLFSR 3\nLFSR 4\nLFSR 1\nXOR\n+\nXOR\n+\n/2\nz–1\nz–1\nT1\nT2\nInitial Values\nx1\nt\nx1\nt\nx2\nt\nx2\nt\nc0\nt\nx3\nt\nx3\nt\nx4\nt\nx4\nt\n1\n2\n2\n2\n2\n2\n2\n3\n3\nct\nyt\nct+1\nSt+1\nEncryption Stream Zt\nSummation Combiner Logic\nBlend\nFigure 11.1\nSchematics of the E 0 encryption engine.\nOn every clock cycle, it moves to a new state ct and produces a single\noutput bit of the key stream Zt. E 0 makes use of four Linear Feedback Shift\nRegisters (LFSR1, . . . , LFSR4) of lengths L 1 = 25, L 2 = 31, L 3 = 33, and\nL 4 = 39 bits with the following feedback polynomials:\nLFSR1 : f1(t) = t25 + t20 + t12 + t8 + 1,\nLFSR2 : f2(t) = t31 + t24 + t16 + t12 + 1,\nLFSR3 : f3(t) = t33 + t28 + t24 + t4 + 1,\nLFSR4 : f4(t) = t39 + t36 + t28 + t4 + 1.\nThe total length of the registers is 128 bits. These primitive polynomials\nhave been chosen as they exhibit the best trade-off between hardware\nimplementation constraints and excellent statistical properties of the output\nsequences (the polynomials are maximum length windmill polynomials\n[13,14]). Let xi\nt denote the tth symbol of LFSRi. The value yt is the sum\nover the integers of the four-tuple x1\nt , x2\nt , x3\nt , x4\nt . Thus yt can take the values\n0, 1, 2, 3, or 4. The output of the summation generator is obtained by the\n" }, { "page_number": 368, "text": "Security in Wireless PAN Mesh Networks\n■\n357\nTable 11.1\nE0 Linear Bijections\nx\nT1[x]\nT2[x]\n00\n00\n00\n01\n01\n11\n10\n10\n01\n11\n11\n10\nfollowing equations:\nZt = x1\nt ⊕x2\nt ⊕x3\nt ⊕x4\nt ⊕c0\nt ,\nSt+1 =\n\u0002S1\nt+1, S0\nt+1\n\u0003 =\n\u0004 yt + ct\n2\n\u0005\n,\nct+1 =\n\u0002c1\nt+1, c0\nt+1\n\u0003 = St+1 ⊕T1[ct] ⊕T2[ct−1],\nwhere T1[.] and T2[.] are two different linear bijections over GF(4), sum-\nmarized in Table 11.1, and c0\nt is the least significant bit of ct. The stream\ncipher E 0 needs to be initialized with the initial values for the four LFSRs\n(altogether 128 bits) and the four bits that specify the values of c0 and c−1.\nThe 132-bit initial value is derived from three inputs: the encryption key\nKC, the Bluetooth hardware address, and the clock of the master (see also\nSection 11.2.2). With the key stream generator, 200 stream cipher bits are\ngenerated, of which the last 128 are fed back into the key stream generator\nas the initial values of the four LFSRs. The values of c0 and c−1 are kept.\nBluetooth makes use of the key derivation algorithms E 1, E 21, E 22, and\nE 3 to map a 128-bit input to a 128-bit output. All of them are based on\nthe SAFER+ block cipher. This is an improved version of the SAFER block\ncipher, which only works on 64-bit data blocks. An important improvement\nin SAFER+ is the introduction of the Armenian Shuffle permutation, which\nboosts the diffusion of single bit modifications in the input data. It is a\npermutation of 16 bytes. SAFER+ consists of:\n■\nA key scheduling algorithm that produces 17 different 128-bit\nsubkeys\n■\n8 identical rounds\n■\nAn output transformation, which is implemented as a bitwise XOR\nbetween the output of the last round and the last subkey\nEach SAFER+ round calculates a 128-bit word out of two subkeys (the last\nsubkey is used in the SAFER+ output transformation) and a 128-bit input\nword from the previous round. The central components of the SAFER+\nround are the 2-2 Pseudo Hadamard Transform (PHT) [15], the Armenian\n" }, { "page_number": 369, "text": "358\n■\nSecurity in Wireless Mesh Networks\nShuffles, and the substitution boxes denoted E and L [11,16]. The PHT\ntakes two input bytes and produces two output bytes, as follows:\nPHT [a, b] = [(2a + b) mod 256, (a + b) mod 256] .\nThe two mappings E and L introduce nonlinearity and are defined as\nfollows:\nE [x] = (45x mod 257) mod 256,\nL [x] = y\nsuch that\nx = E [y].\nThe structure of one SAFER+ round can be found in [17,18]. For a summary\nof recent cryptanalytic results, see [19].\n11.2.2\nKey Agreement Protocol in Bluetooth\nThe Key Agreement Protocol [20] is a crucial part of the security architecture\nof Bluetooth [21]. Suppose that two Bluetooth devices, called A and B,\nwant to communicate securely (in the rest of this chapter, we will assume\nthat A initiates the communication). Initially, these devices do not share a\nsecret. They perform a Key Agreement Protocol to generate a link key and\nan encryption key. The latter is fed to the stream cipher E0. The process\nof generating a shared secret is called pairing (two Bluetooth devices are\npaired when they share a key which can be used to communicate securely).\n11.2.2.1\nGeneration of the Unit Key\nWhen a Bluetooth device is turned on for the first time, it calculates a unit\nkey. This is a key that is unique for every device and that is almost never\nchanged. It is stored in non-volatile memory. The unit key is only used if\none of the devices does not have enough memory to store session keys (see\nalso Section 11.2.2 for more details). The unit key is based on a random\nnumber and the Bluetooth hardware address of the device.\n11.2.2.2\nGeneration of the Initialization Key\nAt the start of a communication session, the Bluetooth devices do not yet\nshare a session key, and will have to establish one. This is achieved in\ndifferent steps. First, an initialization key is generated. This temporary key\nis a function of a random number IN RAND (generated by A and sent to\nB in clear), a shared PIN, and the length L of this PIN. The PIN should\nbe entered in both devices by a user or it can be fed from a higher layer\ninto the pairing procedure. The length of the PIN can be chosen between\n8 and 128 bits. Typically, it consists of four decimal digits. If one of the\ndevices does not have an input interface, a fixed PIN can be used (often,\n" }, { "page_number": 370, "text": "Security in Wireless PAN Mesh Networks\n■\n359\nB\nA\nE22\nL\nE22\nIN_RAND\nIN_RAND\nIN_RAND\nPIN\nL\nPIN\nKinit\nKinit\nFigure 11.2\nGeneration of the initialization key.\nthe default value is 0000). This procedure is shown in Figure 11.2. The\nresult is a temporary shared key: the initialization key. Note that a low-\nentropy shared secret (the PIN) is used to generate the initialization key.\nAs a consequence, an eavesdropper, which is present during initialization,\nwill know the random number IN RAND.\n11.2.2.3\nMutual Entity Authentication\nEach time a new shared key is generated (an initialization key or a link\nkey), both devices perform a mutual authentication protocol. The authenti-\ncation scheme is based on a challenge-response protocol. This protocol is\nperformed twice. First, B authenticates itself to A, as shown in Figure 11.3.\nIf this authentication is successful, the roles are switched (B becomes the\nverifier and A the prover). The authentication goes as follows. A gen-\nerates a random number AU RAND and sends this to B. This random\nnumber is called the challenge. Both devices now compute a response\nSRES = E 1(ADDRB, Klink, AU RAND). ADDRB is the Bluetooth hardware\naddress of B and Klink is the shared key (initialization key or link key). B\nsends its response to A. If this response corresponds to the value that A has\ncalculated, then the authentication is successful. The value ACO (Authen-\nticated Ciphering Offset) is used for the generation of the encryption key.\nSRES\nB\nA\nADDRB\nE 1\nSRES\nACO\nKlink\nAU_RAND\nADDRB\nKlink\nAU_RAND\nAU_RAND\nE 1\nSRES\nACO\nE 1\nE 1\nFigure 11.3\nMutual entity authentication protocol.\n" }, { "page_number": 371, "text": "360\n■\nSecurity in Wireless Mesh Networks\nA\nADDRA\nLK_RANDA\nEKinit(LK_RANDB)\nEKinit(LK_RANDA)\nADDRB\nLK_RANDB\nLK_KA\nLK_KB\nKAB = Klink\nE21\nE21\nB\nADDRB\nLK_RANDB\nADDRA\nLK_RANDA\nLK_KB\nLK_KA\nKAB = Klink\nE21\nE21\nFigure 11.4\nThe link key is a combination key.\nAlgorithm E 1 is based on the SAFER+ block cipher, with some small\nmodifications [11].\n11.2.2.4\nGeneration of the Link Key\nBoth devices now share an initialization key. This key will be used to\nagree on a new, semi-permanent key (called the link key). The link key\nwill be stored on both devices for future communication. Depending on\nthe memory constraints of both devices, the link key can be the unit key\nof the memory-constrained device or a combination key derived from the\ninput of both devices (Figure 11.4).\nIf the unit key of device A is the link key, it is transmitted encrypted\nfrom A to B. This encryption is done by XORing the unit key of A with\nthe initialization key.\nIf the link key is a combination key, then both devices first generate\na random number LK RAND. These random numbers are encrypted with\nthe initialization key and sent to the other device. Now they both compute\nLK KA = E 21(LK RANDA, ADDRA) and LK KB = E 21(LK RANDB, ADDRB).\nThe combination key K AB is the XOR of LK KA and LK KB. This is shown\nin Figure 11.4. Algorithm E 21 is based on the SAFER+ block cipher, with\nsome small modifications. After the generation of the link key, the (old)\ninitialization key is definitively discarded and a mutual authentication is\nstarted, using the exchanged link key that is shared between both devices\n(this has already been discussed). The procedure shown in Figure 11.4 is\nalso carried out when a new link key is computed. The only difference is\nthat the random numbers LK RAND are encrypted with the old link key.\nAfter the generation of the new link key, the old one will be discarded.\n" }, { "page_number": 372, "text": "Security in Wireless PAN Mesh Networks\n■\n361\nCOF\nE3\nKC\nKlink\nEN_RANDA\nEN_RANDA\nK’C\nB\nA\nRED\nCOF\nE3\nKC\nKlink\nEN_RANDA\nK’C\nRED\nFigure 11.5\nGeneration of the encryption key.\n11.2.2.5\nGeneration of the Encryption Key and the Key Stream\nAfter a successful generation of the link key and execution of the mutual\nauthentication protocol, the encryption key can be generated. Device A\ngenerates a random number EN RANDA and sends this to B. Both devices\ngenerate the encryption key KC = E 3(EN RANDA, Klink, COF). The COF\nvalue (Ciphering Offset Number) is the ACO value which was generated\nduring the mutual authentication protocol. However, if the encryption key\nis used for broadcast, then the COF is the concatenation (denoted by ||) of\nthe Bluetooth hardware address ADDR of the sender and itself (so COF =\n(ADDR || ADDR)). The encryption key KC has a length of 128 bits, but its\nlength can be reduced to a truncated encryption key K ′\nC if necessary. This\nprocedure is shown in Figure 11.5.\nFinally, the encryption key KC (or the truncated key K ′\nC) is fed to the\nencryption scheme E 0 together with the Bluetooth hardware address and\nthe clock of the master. These values are used to initialize the four LFSRs\nof the stream cipher E 0. The output of the cipher is the key stream Kcipher\n(see Figure 11.6). The master clock is used to make the key stream harder\nto guess.\n11.2.3\nSecurity Weaknesses in the Bluetooth Security\nArchitecture\nThere are several security weaknesses in the Bluetooth standard [21,22].\nWe now give an overview of the most important security problems.\n" }, { "page_number": 373, "text": "362\n■\nSecurity in Wireless Mesh Networks\nB\nA\nADDRA\nE0\nKcipher\nK’C\nclockA\nADDRA\nE0\nK’C\nclockA\nKcipher\nFigure 11.6\nGeneration of the key stream.\n11.2.3.1\nUnit Key\nThe unit key is employed if one of the Bluetooth devices does not have\nenough memory to store session keys. This key is stored in non-volatile\nmemory and almost never changed. As already described in Section 11.2.2,\nthe unit key is sent encrypted (with the initialization key) to the other\ndevice. The result is the following weakness: if A has sent its unit key to\ndevice B, then B knows the key of A and can impersonate itself as A to\na device C. This impersonation attack is impossible to detect. It is strongly\nrecommended to avoid the use of unit keys!\n11.2.3.2\nLocation Privacy\nWhen two or more Bluetooth devices are communicating, the transmit-\nted packets always contain the Bluetooth hardware address of the sender\nand the destination (or an identifier which is directly related to these ad-\ndresses). When an attacker eavesdrops on the transmitted data, he knows\nthe Bluetooth addresses of these devices. The attacker does not have to be\nphysically close to the communicating devices, he can use a device with\na stronger antenna (e.g., it is very easy to construct an antenna which can\nintercept Bluetooth communication from more than one mile away [23,24])\nor just place a small tracking device near the two Bluetooth devices.\nThis way, the attacker can keep track of the place and time these de-\nvices were communicating. This is a violation of the privacy of the user. The\nlocation information can be sold to other persons or used for location de-\npendent commercial advertisements (e.g., a shop can send advertisements\nto everybody that is near the shop). It should be possible for the user to\ndecide when his location is revealed and when not.\n11.2.3.3\nSecurity Depends on Security of PIN\nThe initialization key is a function of a random number IN RAND, a shared\nPIN, and the length L of the PIN. The random number is sent in clear and\n" }, { "page_number": 374, "text": "Security in Wireless PAN Mesh Networks\n■\n363\nhence known by an attacker who is eavesdropping during the initialization\nphase. This means that only the PIN is unknown to the attacker. If an\nattacker obtains the PIN, he knows the initialization key. Worse yet, because\nall the other keys are derived from the initialization key, they will also be\nknown by the attacker. Hence the security of the keys used in Bluetooth\ndepends on the security of the PIN. If this value is too short or weak\n(e.g., 0000), it is very easy for an attacker to guess the PIN (and hence\nthe initialization key). Unfortunately, it is very cumbersome for a user to\nremember long (and random) numbers.\nNote that it is possible to verify a guess of the PIN. The reason is that a\nmutual authentication protocol is executed after the generation of the ini-\ntialization key. If an attacker observes this protocol, he obtains a challenge\nand the corresponding response. The attacker calculates for every guess of\nthe PIN the corresponding response and when this is equal to the observed\nresponse, the guess of the PIN was correct. The shorter the PIN, the faster\nthis brute-force attack can be carried out. Shaked and Wool showed that\nthis attack can be optimized by employing an algebraic representation of\nSAFER+, the cryptographic primitive used in the mutual authentication pro-\ntocol [16] (see Section 11.2.1). The authors state that a PIN of four digits\ncan be cracked in less than 0.06 seconds on a standard PC. This is a very\ncritical security problem.\n11.2.3.4\nDenial-of-Service Attacks\nMobile networks are always vulnerable to denial-of-service (DoS) attacks.\nThey consist of mobile devices, and these devices are often battery pow-\nered. Bluetooth is no exception. An attacker can send dummy messages to\na mobile device. When this device receives a message, it performs some\ncomputations, which consumes battery power [25]. After some time, all\nbattery power will be consumed. This exhaustion of the battery power is\ncalled the sleep deprivation attack [26]. This attack is almost impossible to\nprevent.\nThere are also some more advanced DoS attacks, caused by implemen-\ntation decisions. A nice example is the black list, which is used during the\nmutual authentication protocol. To avoid that a device would start the\nauthentication protocol over and over again (and eventually guess the cor-\nrect PIN), each device has a black list of the Bluetooth addresses of the\ndevices which failed to authenticate themselves correctly. These devices\ncannot start an authentication procedure during some period. Each con-\nsecutive time the authentication procedure fails, this period is increased\nexponentially (until a pre-determined upper limit is reached). Candolin\ndiscovered that this mechanism can be exploited in several DoS attacks\n[26]. An attacker can try to authenticate itself to device A, but change its\nBluetooth hardware address every time. All these authentication attempts\n" }, { "page_number": 375, "text": "364\n■\nSecurity in Wireless Mesh Networks\nwill fail and the black list of A will become quite large. If there is no upper\nlimit on this black list, the entire memory of A will be filled with the entries\nof the black list and device A will crash.\nThis is not the only DoS attack. Suppose device B wants to authenticate\nitself to A. After A has sent a challenge to B, the attacker sends a wrong re-\nsponse to A using the Bluetooth hardware address of B. The authentication\nwill fail, B will be put on the black list of A, and the (correct) response of\nB will be ignored by A. The attacker keeps repeating this attack and B will\nnever be able to authenticate itself successfully to A. Note that the same\nresult could be obtained by jamming the radio signal, but the DoS attacks\ndescribed above are much easier to perform.\n11.2.3.5\nEncryption Algorithm E0\nBluetooth uses the stream cipher E 0 for data encryption. This stream cipher\nhas some security flaws [27–32]; note though that most of the published\nattacks do not work on the implementation of E 0 in Bluetooth.\nThe attacks with the lowest complexity are the algebraic attacks [28].\nE 0 is vulnerable to algebraic attacks because of the possibility to recover\nthe initial value by solving a system of non-linear equations of degree 4\nover the finite field GF (2). This system can be transformed by linearization\ninto a system of linear independent equations with at most 223 unknowns.\nFortunately, this attack does not work in Bluetooth because it needs a long\nkey stream during the initialization and E 0 in Bluetooth only uses small\npackets (the payload ranges from zero to a maximum of 2745 bits [4]).\nThere are, however, some attacks which can be implemented on the\nE 0 algorithm in Bluetooth. Most of them are not very efficient, but re-\ncently Vaudenay found a practical known-plaintext attack [33]. This is the\nfastest attack on the Bluetooth encryption scheme. The attack is based on\na recently detected flaw in the resynchronization of E 0, as well as the in-\nvestigation of conditional correlations in the FSM governing the keystream\noutput of E 0. This attack finds the original encryption key for two-level E 0\nusing the first 24 bits of 223.8 frames, requiring 238 computations.\n11.2.3.6\nBluejacking\nWhen two Bluetooth devices are paired, these devices will send their\n“name” to each other. The default name of a device is typically the brand\nname (e.g., “NOKIA 6110”). The user can, however, change this name in\nan arbitrary string (up to 248 characters) and this user-defined name will\nbe displayed on the output interface of the other device. The goal of this\nname is to facilitate the pairing process. First, the device displays a list of\nall the names of the discoverable devices in the neighborhood. The user\nthen selects the name of the device that it wants to pair its device with. The\nBluejacking attack [34] exploits this name to send advertisements to other\n" }, { "page_number": 376, "text": "Security in Wireless PAN Mesh Networks\n■\n365\nBluetooth devices. The name of the malicious sender is the advertisement\nitself (e.g., “buy product X now”). A malicious user can try to start a pair-\ning process with all the discoverable devices in the neighborhood and this\nforces its name to be displayed on the other devices. This is not really a\ncritical security problem, but it can become annoying (e.g., think of the\namount of SPAM e-mails a user receives daily). By choosing a misleading\nname, a malicious device could try to force a pairing process with another\ndevice.\n11.2.3.7\nImplementation Errors\nImplementation errors can result in critical security problems. A good ex-\nample is the Bluesnarf attack [35]. It is possible, on some mobile phones,\nto connect to the device without alerting the owner of the target device\nof the request, and gain access to restricted portions of the stored data in\nthe phone, including the entire phone book (and any image or other data\nassociated with the entries), calendar, real-time clock, business card, prop-\nerties, change log, IMEI (International Mobile Equipment Identity, which\nuniquely identifies the phone to the mobile network, and is used in ille-\ngal “phone cloning”), etc. This is normally only possible if the device is in\ndiscoverable mode, but there are tools available that allow even this safety\nnet to be bypassed.\nThe Bluesnarf attack can also be extended by combining it with a back-\ndoor attack [35]. The result of this combined attack is that not only the\nprivate data of the mobile phone can data be retrieved, but other services\nsuch as access to the Internet, WAP [36], and GPRS gateways or even send-\ning an SMS are available for the attacker without the owner’s knowledge.\nThese attacks are caused by implementation errors and hence can be fixed\nby the vendors.\n11.2.3.8\nOther Security Problems\nThere are also some security problems in the challenge-response protocol,\nwhich uses the algorithm E 1 and is based on the SAFER+ block cipher.\nKelsey et al. [37] discovered a weakness in the key schedule of SAFER+\nthat allows a key search to be performed slightly faster than by exhaustive\nsearch. This attack is only a theoretical issue and does not really endanger\nthe security of Bluetooth. But it indicates that it would be better to replace\nthe SAFER+ block cipher by, for example, AES.\nAnother security flaw is the lack of integrity checks on the Bluetooth\npackets. An attacker can always modify a transmitted Bluetooth packet\nwithout being detected. Note that encryption in itself does not offer any\nintegrity protection.\nMan-in-the-middle attacks are also not prevented in Bluetooth. The rea-\nson is that the data is never authenticated by the sender. And there are\n" }, { "page_number": 377, "text": "366\n■\nSecurity in Wireless Mesh Networks\nalmost no time stamps or nonces in the protocols, so the freshness of the\nmessages is not guaranteed. Suppose that an attacker has obtained a link\nkey used by two devices. The attacker can now establish a new link with\neach of the devices, pretending to be the other device. The two devices\nstill believe that they are talking to each other, but in fact they are commu-\nnicating with the attacker.\nTo make things even worse, a user can switch off security. Often, the\ndefault configuration is no security at all. This certainly has to be avoided.\n11.2.4\nBluetooth Security in Practice\nAlthough there are several security problems in the Bluetooth standard, it\nis certainly possible to use Bluetooth in security-critical applications. Here\nare some recommendations for designers of Bluetooth applications:\n■\nAvoid the use of unit keys, as this will jeopardize the security.\n■\nProvide data integrity protection in one of the layers on top of Blue-\ntooth. This means that the integrity of the payload cannot be checked\nin the MAC layer, and that the received data has to be passed to the\nhigher layer. This is, however, still a lot better than no data integrity\nprotection at all.\n■\nIf one uses IP over Bluetooth, and the mobile devices are not energy\nconstrained (e.g., a laptop), one can employ standardized solutions\nlike IPSec to protect the security of the Bluetooth link.\n■\nIn all the other scenarios, one can implement an advanced pairing\nprotocol [38–41] to securely establish a session key between the\nmobile devices that want to communicate.\n■\nThe use of pseudonyms can make the system robust against tracking.\nThis requires, however, a modification of the Bluetooth standard or\nspecialized hardware.\n■\nFinally, make sure that security is always turned on, certainly in\nthe default configuration (as users tend to use this configuration the\nmost).\n11.3\nZigBee Security\nZigBee is a set of communication protocols that operate on the application\n(APL) and network (NWK) layer. It works on top of the low-power MAC\nand PHY layer, which are standardized in the IEEE 802.15.4 standard for\nWPANs. One of the design principles of ZigBee is that the layer that origi-\nnates a frame is responsible for securing it. So, if an NWK command frame\nneeds protection, NWK layer security shall be employed. Figure 11.7 shows\nan example of the security fields that may be included in an NWK frame.\n" }, { "page_number": 378, "text": "Security in Wireless PAN Mesh Networks\n■\n367\nX1\nE\nE\nE\nX2\nX3\nB0\nB1\nB2\nk\nk\nk\n0\n...\nFigure 11.7\n(Part of) ZigBee frame with security at the NWK level.\nThe auxiliary header contains security information (security control, frame\ncounter, etc.), the payload can be encrypted or not, and the Message In-\ntegrity Code (MIC) is used to protect the integrity of both header fields and\nthe payload (the security control field in the auxiliary header specifies the\nlevel of security that is applied to the frame). Both encryption and mes-\nsage integrity are provided by one building block: the CCM ∗algorithm.\nSecurity information is stored in Access Control Lists (ACLs). Each ACL\nentry contains the following security information: destination address, se-\ncurity control field, key, nonce, and the key and frame counter. The frame\ncounter is incremented by one for every outgoing frame. The maximum\nvalue is 232 −1. When a new key is used, the frame counter is reset to 0.\nThere is always a default ACL entry which is used if there is no specific\nACL entry for the destination. There can be maximally 255 ACL entries. The\nexact amount of ACL entries is vendor specific.\nZigBee uses the open trust model [6]. This implies that all different lay-\ners of the communication stack, and all applications running on a single\ndevice, trust each other. Keys can be reused in each layer. To simplify in-\nteroperability, the security level used by all devices in a given network and\nby all layers of a device shall be the same. If protection from theft of service\nis required, NWK layer security shall be used for all frames. The network\nkey (NWK key) is a broadcast key that is used by all devices in the same\nnetwork. As a consequence, using an NWK key does not prevent insider\nattacks. The NWK key is updated regularly and is stored in the default ACL\nentry. To distinguish between the different NWK keys and to make sure that\nevery device in the network is using the most recent NWK key, a sequence\nnumber (called the key counter) is assigned to every NWK key. The NWK\nkey is only used in the NWK layer. If application layer security is applied,\na link key is used to protect outgoing frames. Link keys are employed to\nenable end-to-end security (between source and destination device).\n" }, { "page_number": 379, "text": "368\n■\nSecurity in Wireless Mesh Networks\n11.3.1\nZigBee Cryptographic Primitives\n11.3.1.1\nCCM ∗Algorithm\nCCM ∗is a generic combined encryption and authentication block cipher\nmode. CCM ∗is only defined for use with block ciphers with a 128-bit block\nsize. The block cipher that is used in the ZigBee specification is the AES-\n128. The CCM ∗mode is a minor modification of the CCM mode specified\nin the IEEE 802.15.4 MAC layer specification [7]. CCM ∗includes all of the\nfeatures of CCM and additionally offers encryption-only and integrity-only\ncapabilities. In total, there are eight possible security levels: the payload of\na frame can be encrypted or not, and the length of the MIC, which protects\nthe integrity of the header fields and the payload of a frame, can be 0,\n32, 64, or 128 bits. The security control field in the header specifies which\nsecurity level is used to secure the frame. As the CCM mode, the CCM ∗\nmode requires only one 128-bit key. Together with this key, a unique 104-\nbit nonce N is used. This nonce is a function of the security control field,\nthe frame counter, and the address of the sender. Within the scope of a key,\nthe nonce value should be unique. The frame counter prevents reusing a\nnonce under the same key.\nAn authentication tag T is computed as follows (see also Figure 11.8):\nT = Xt+1,\nXi+1 = E (key, Xi ⊕Bi)\nfor\ni = 0, . . . , t .\nE is the block cipher AES-128, B1∥. . . ∥Bt are the t data blocks that have to\nbe integrity protected (each block has a length 128 bits), B0 is a data block\nthat contains the nonce N and some constants, and X0 is a 128-bit block\ncontaining only 0s. The authentication tag T holds the M left-most bits of\nthe output Xt+1. The value M specifies the length (in bytes) of the MIC.\nNote that the block cipher is used in Cipher-Block Chaining (CBC) mode\n[10].\nNWK\nHeader\nAuxiliary\nHeader\n(Encrypted) NWK\nPayload\nMIC\nFigure 11.8\nCCM ∗authentication block cipher mode.\n" }, { "page_number": 380, "text": "Security in Wireless PAN Mesh Networks\n■\n369\nC1\nE\nA1\nk\n...\nM1\nC2\nE\nA2\nk\nM2\nC3\nE\nA3\nk\nM3\nFigure 11.9\nCCM ∗encryption block cipher mode.\nEncryption is performed as follows:\nAi = Flags∥N∥i\nfor\ni = 1, . . . , t ,\n(11.1)\nCi = E (key, Ai) ⊕Mi\nfor\ni = 1, . . . , t ,\n(11.2)\nS0 = E (key, A0) .\nFirst, the 128-bit blocks Ai are computed. They contain the constant value\nFlags (8-bit representation of the value 1), the nonce N, and a 16-bit counter\ni. These blocks are fed to the block cipher AES-128. The output is XORed\nwith the t data blocks Mi that have to be encrypted (each block has a\nlength of 128 bits), and the result is the t cipher text blocks Ci (see also\nFigure 11.9). The M left-most bits of block S0 are XORed with the authen-\ntication tag T . The result is the encrypted authentication tag U . The MIC\nis equal to T or U (depending on if encryption is applied or not), and the\nencrypted payload to C1∥. . . ∥Ct.\n11.3.1.2\nThe AES Algorithm\nAES is a symmetric block cipher with a block-length of 128 bits and three\ndifferent key sizes: 128, 192, and 256 bits. The three resulting algorithms\nare referred to as AES-128, AES-192, and AES-256. The cipher is based on\na round operation that is repeated a number of times. Each round has two\ninputs: a round-key of 128 bits and the result of the previous round. The\nround-keys can be pre-computed or generated on-the-fly out of the input\nkey. Every round consists of four steps: Byte Substitution, Shift Rows, Shift\nColumns, and Add Round Key (this simply XORs the round-key with the\ncurrent block). The number of rounds depends on the size of the key: 9,\n11, and 13 rounds for 128-, 192-, and 256-bit keys, respectively. Due to its\nregular structure, AES can be implemented very efficiently in hardware and\n" }, { "page_number": 381, "text": "370\n■\nSecurity in Wireless Mesh Networks\nsoftware. Computational performance of software implementations often\ndiffers between encryption and decryption because the inverse operations\nin the round function are more complex than the according operation for\nencryption. For further information, we refer to [8].\n11.3.2\nSecurity Architecture of ZigBee\n11.3.2.1\nKey Hierarchy\nSeveral types of keys are used in ZigBee, forming a key hierarchy. Typically,\nthe security manager of a device (situated in the application layer) will\nperform the following steps:\n1.\nObtain the trust center master key: Initially, each device shares a\ntrust center master key with the trust center. The device can obtain\nthis trust center master key (together with the address of the trust\ncenter) in two ways: the device acquires the trust center master\nkey via insecure key-transport (e.g., it is sent in clear from the trust\ncenter to the device at low power) or it acquires this key via pre-\ninstallation (e.g., factory installation or based upon data entered by\na user). It is very important that no other device can obtain this trust\ncenter master key, as the security of all other keys used in ZigBee\ndepends on the confidentiality of the trust center master key.\n2.\nEstablish link key with trust center: The trust center and the device\nshare a trust center master key and will execute the Symmetric-\nKey Authenticated Key Agreement (SKKE) protocol to establish a\nlink key with each other. First, both devices generate a random\n128-bit challenge (QEU and QEV, respectively) and send it to the\nother device. These challenges are fed, together with the trust center\nmaster key, to a key derivation function. The result is two 128-bit\nkeys: the MacKey and the KeyData. The former is the key of an\nMIC, used to mutually authenticate the challenges QEU and QEV.\nAfter a successful authentication, both devices will use the KeyData\nkey as shared link key. This link key will be employed to secure\nthe communication between the trust center and the device.\n3.\nCompute key-load key: The key-load key is derived from the link\nkey as follows:\nkey-load key = HMAClink key(0 × 02) .\nHere, HMAC is a keyed message authentication code [10]. This type\nof MAC function uses a cryptographic hash function in combination\nwith a secret key. The trust center uses the key-load key to transport\nan application master key securely to a device.\n" }, { "page_number": 382, "text": "Security in Wireless PAN Mesh Networks\n■\n371\n4.\nCompute key-transport key: The key-transport key is derived from\nthe link key as follows:\nkey-transport key = HMAClink key(0 × 00) .\nThe trust center uses the key-transport key to transport an applica-\ntion link key or an NWK key securely to a device.\n5.\nObtain the NWK key: The trust center puts the NWK key (that is\ncurrently being used in the network) in a specially constructed com-\nmand frame, secures it with the key-transport key, and transmits it\nto the device. The NWK key is used to encrypt broadcast com-\nmunication in the network. Note that command frames are always\nencrypted and integrity protected (with a 128-bit MIC).\n6.\nObtain the application link key: When two devices in a network\nwant to communicate securely (end-to-end), they need an appli-\ncation link key. One way to obtain such an application key is as\nfollows: the trust center generates the application link key and puts\nit in a specially constructed command frame. This frame is sent\nsecurely to each device. The security of the frame is protected by\nemploying the key-transport key. The advantage of the trust center\nsending out the application link keys directly is that key-escrow can\nbe implemented.\na.\nObtain the application master key: Instead of directly transmit-\nting the application link key to both devices, the trust center\ncan also generate an application master key. It puts this key\nin a specially constructed command frame, and sends this se-\ncurely to both devices. The security of this frame is protected\nby employing the key-load key.\nb.\nEstablish application link key with other devices: After the de-\nvices obtained the application master key, they execute the\nSKKE protocol. This is done exactly as described above. The\nonly difference is that the application master key is used to\nderive the link key, instead of the trust center master key. The\noutput of the SKKE protocol is the application link key, which\nis used for end-to-end security between both devices.\nThe above is only valid if the trust center is working in commercial mode.\nWhen the trust center works in residential mode, the device will not estab-\nlish a link key with other devices. A more detailed discussion on the modes\nof operation of the ZigBee trust center is now presented.\n11.3.2.2\nZigBee Trust Center\nThere is always exactly one trust center in each secure ZigBee network.\nThis device is often the ZigBee coordinator and is trusted by all devices in\n" }, { "page_number": 383, "text": "372\n■\nSecurity in Wireless Mesh Networks\nthe network. It is responsible for the distribution of keys (link keys and\nNWK keys) among the ZigBee devices. The ZigBee trust center also en-\nforces the policies in the network. These policies state how a device can\njoin or leave the network (securely or insecurely), if and when keys have\nto be updated, etc. The trust center can be configured to operate in either\ncommercial or residential mode:\n■\nThe commercial mode of the trust center is designed for high-security\ncommercial applications. In this mode, the trust center maintains a\nlist of devices, master keys, application link keys, and NWK keys that\nit needs to control. It also enforces the policies of NWK key updates\nand network admittance. In this mode, the memory required for the\ntrust center grows with the number of devices in the network. When\nthe trust center works in commercial mode, it shall follow the steps\nof the key hierarchy described above.\n■\nThe residential mode of the trust center is designed for low-security\nresidential applications. In this mode, the trust center maintains a list\nof the NWK keys and controls the policies of network admittance. It\ndoes not have to maintain a list of devices, master keys, or applica-\ntion link keys. When operating in residential mode, the NWK key is\nnever updated, and therefore the memory required for the trust cen-\nter does not grow with the number of devices in the network. This\nlimits the implementation complexity, but also reduces the security.\nWhen the trust center works in residential mode, it shall not follow\nthe steps of the key hierarchy described above. Instead, it will just\nsend the NWK key to a device joining the network via insecure key\ntransport. This key is used to secure communication. Master keys\nand link keys are not employed.\n11.3.3\nSecurity Weaknesses in the ZigBee Security\nArchitecture\nImproper use of the security mechanisms in ZigBee can cause several se-\ncurity problems [42,43]. ZigBee has, however, solved some security issues\nthat were present in the IEEE 802.15.4 standard [6], e.g., limiting the en-\ncryption mode to CCM ∗in ZigBee avoids the employment of dangerous\nsecurity modes, like AES-CTR. We now give an overview of the most im-\nportant security problems that still remain in ZigBee. Designers of ZigBee\napplications should take this into account during implementation.\n11.3.3.1\nIV (Nonce) Management Problems\nAs already discussed in the previous section, security information is stored\nin ACLs. Each ACL entry contains the following security information: des-\ntination address, security control field, key, nonce, and the key and frame\n" }, { "page_number": 384, "text": "Security in Wireless PAN Mesh Networks\n■\n373\ncounters. The nonce is a function of the security control field, the frame\ncounter, and the address of the sender. Only the frame counter is really vari-\nable, and as a consequence, the nonce is derived directly from the frame\ncounter. Suppose one would encrypt two messages (M1 and M2) with\nthe same key and the same nonce. According to Equation 11.1, reusing\na nonce results in reusing the block Ai. If we apply Equation 11.2, one\nobtains the following result:\nC1 ⊕C2 = E (key, Ai) ⊕M1 ⊕E (key, Ai) ⊕M2 = M1 ⊕M2 .\nThis should certainly be avoided! Fortunately, the frame counter prevents\nreusing a nonce under the same key. There is, however, a problem if a key\nis used in two different ACLs (because in this case, the frame counter in\neach ACL is updated independently and this could result in the reuse of\na nonce) or if a nonce is reused in the same ACL (without the key being\nupdated). The latter can occur when a power failure arises. If the frame\ncounter is stored in volatile memory, and the key in non-volatile memory,\nthen the frame counter would be reset to zero after the power failure. The\nkey, however, would remain the same, and one would reuse the nonce\nunder the same key. To avoid this problem, the frame counter and the\nkey should be stored together in non-volatile memory. The same problem\nwould occur if one would use a key that has been employed before, but\nthe probability of such an event to occur is very low.\n11.3.3.2\nImproper Support of Group Keying\nZigBee does not support group keying. The reason is that each ACL can\nonly contain the address of one destination. Let us assume that one would\nuse multiple ACLs, one for each destination in the group. Then the proba-\nbility of reusing a nonce would become very large. As explained above, a\nnonce should never be reused under the same key. If one would use one\nACL for the entire group, then one always has to update the address of\nthe destination beforehand (otherwise, the device cannot find the correct\nACL entry in its memory). This is not possible, because one would have\nto know in advance which device is going to send the next message, and\nnormally a device does not have this knowledge. Another problem would\nbe that each device in the group has to update the frame counter every\ntime a message is sent to one of the group members, also when it was not\nintended for the device itself. So ZigBee only supports secure unicast and\nbroadcast communication, and no secure multicast communication.\n11.3.3.3\nKey Management\nThe ZigBee standard states that there can be maximally 255 ACL entries.\nThe exact amount of ACL entries is vendor specific and often much lower\n" }, { "page_number": 385, "text": "374\n■\nSecurity in Wireless Mesh Networks\nthan 255. As an example, the Chipcon CC2420 has support for only two ACL\nentries [43]. The number of application link keys a device can maximally\nshare with other devices is equal to the number of ACL entries. So in the\nbest case, it can only share a key with 255 other ZigBee devices, which is\nconsiderably less than the maximum amount of 65,536 devices in a ZigBee\nnetwork. A better support for secure end-to-end communication is needed.\n11.3.3.4\nReplay Attacks\nEvery time a message is transmitted to another device, the frame counter is\nincremented by one. This prevents replay attacks, as frames with a lower\nframe counter than stored in the ACL will be discarded. This can, however,\ncause a security problem in broadcast communication. In a ZigBee network,\nbroadcast communication is secured with the NWK key, which is stored in\nthe default ACL. Every time a message is broadcasted, each device in the\nnetwork should increment the frame counter in its default ACL. If a device\ngoes to sleep mode and does not receive broadcast messages for a certain\ntime, it cannot send any broadcast message anymore. The frame counter\nin its default ACL will have a lower value than the one in the default ACL\nof the other devices, and a message with a lower frame counter will be\ndiscarded by the other devices, as they wrongfully detect this event as a\nreplay attack. As a consequence, a device can never go to sleep mode, and\nthis can have an important influence on the battery lifetime of a ZigBee\ndevice. Requiring each device in the network to update its frame counter\nregularly causes some key management problems and is not very practical.\nIt would be better not to increment the frame counter in case of broadcast\ncommunication, but this would enable replay attacks.\n11.3.3.5\nInitialization Procedure\nThe secure initialization and installation of the master key determines the\nsecurity of the other keys. When an attacker obtains the trust center master\nkey, this would compromise the security of the other keys used in ZigBee,\nas they are all derived from the trust center master key.\nA device can obtain the trust center master key (and the address of the\ntrust center) in two ways: via insecure key-transport or via pre-installation.\nThe former is the easiest method, but also the most insecure one. Trans-\nmitting a key at low power, as suggested in the ZigBee standard, does\nnot provide sufficient protection. The attacker can build a ZigBee device\nwith a strong directional antenna and intercept communication from a\nlong distance. Assuming that there is no attacker present during the inse-\ncure key-transport is a very dangerous assumption. Theoretically, insecure\nkey-transport is only secure when it is conducted in a Faraday cage. This\nis, however, not very practical. That is why it is recommended to obtain\nthe trust center master key via pre-installation. This is more awkward, but\n" }, { "page_number": 386, "text": "Security in Wireless PAN Mesh Networks\n■\n375\nprovides more security. For example, one could install the trust center ad-\ndress and master key during the fabrication of the ZigBee device. There are,\nhowever, some practical problems. One does not always know in advance\nin which network the ZigBee device will be employed. Deriving the trust\ncenter master key from data entered by a user (a password) can be dan-\ngerous. Users tend to use low-entropy passwords, and an attacker can try\nall passwords or perform a dictionary attack. Because the SKKE protocol,\nused to establish a link key, contains a key confirmation step, an attacker\ncan easily verify every guess of the password.\nThat is why ZigBee needs a secure initialization procedure (e.g., install\nthe keying information via out-of-band mechanisms [38–41,44]). This is a\ncritical security problem that has yet to be solved.\n11.3.3.6\nLocation Privacy\nThe header of a ZigBee frame, which is never encrypted, contains the\naddress of the source and destination device. This address is either the\n64-bit IEEE address, or a 16-bit short address (used once the network is\nset up). When an attacker eavesdrops on the transmitted data, he knows\nthe addresses of the devices that were communicating. It is possible for an\nattacker to construct a stronger antenna to intercept ZigBee communication\nfrom a further distance. As a consequence, an eavesdropper does not have\nto be physically close to the communicating devices.\nThis way, the attacker can keep track of the place and time that ZigBee\ndevices are communicating. This is a violation of privacy. The problem,\nhowever, is less critical than in Bluetooth. In contrast to Bluetooth devices,\nZigBee devices do not always belong to a specific user, but are usually\nused in small sensor networks. In that case, information about the place\nand time a ZigBee device is communicating might not be very interesting\nfor an attacker.\n11.3.3.7\nInsufficient Integrity Protection\nIn total, there are eight security levels that can be employed to secure a\nframe. The payload can be encrypted or not, and the frame can contain\nan MIC of 0, 32, 64, or 128 bits. As a consequence, it is possible to apply\nencryption and no integrity protection on a frame. This is a dangerous\nmode of security and should never be used. Encryption in itself does not\nprovide integrity protection. As shown in Equation 11.2, the cipher text\nCi is the XOR of the plaintext message Mi and the encryption of a block\nAi. This means that if the attacker changes the jth bit of Ci, the same bit\nwill change in the message Mi. This can have important consequences.\nFortunately, the ZigBee standard states that all ZigBee command frames\nshould be encrypted and integrity protected with a 128-bit MIC.\n" }, { "page_number": 387, "text": "376\n■\nSecurity in Wireless Mesh Networks\n11.4\nConclusion and Open Issues\nWe have evaluated the security architectures of both the Bluetooth and\nZigBee standards. We can conclude that both Bluetooth and ZigBee have\nsome (minor) security weaknesses. However, it is still possible to use these\nsystems in a secure way, if the necessary precautions are taken. The security\nweaknesses in Bluetooth range from design problems (e.g., the use of\nunit keys) to problems with the cryptographic algorithms that are used\n(e.g., weaknesses in the E 0 and SAFER+ ciphers). Many of the problems\ncan be mitigated using some practical guidelines (see Section 11.2.4). The\nproblems with the cryptographic ciphers can only be solved by replacing\nthese ciphers or by “patching” them, for example, by switching keys before\nan adversary has enough data to determine the key. ZigBee already solves\na number of the security problems of IEEE 802.15.4 by only allowing the\nCCM∗mode, but still has a number of security problems that should be\nsolved in the next version of the standard.\nThe main difference between the Bluetooth and ZigBee security ar-\nchitectures is that Bluetooth is limited to the MAC layer, but the ZigBee\nstandard also includes the application layer. This results in the fact that\nBluetooth only allows the establishment of link keys between two nodes\nthat are within range, but ZigBee allows any two nodes to establish a shared\nkey. Therefore, ZigBee is more tailored toward wireless mesh networks\nthan Bluetooth.\nOne important issue that has not been solved by either Bluetooth or\nZigBee is location privacy. Both standards allow an adversary to track the\nlocation of devices using the unique identity of the source that is included\nin every frame. To solve this, advanced solutions are required that hide the\nidentity of the devices by employing one-time pseudonyms instead of the\nfixed identifiers.\nA second important open issue is how to securely initialize the security\nmechanisms that are available in a WPAN. Bluetooth only offers the use of a\nPIN that has to be manually entered by the user. One potential solution here\ncould be the use of more advanced pairing protocols. For large scale ad\nhoc networks such as ZigBee, initializing the security mechanisms is even\nharder. An ideal initialization procedure should be very efficient (meaning\nthat extensive use of public key cryptography should be avoided), user\nfriendly (no or very limited user interaction required), and flexible to many\ndifferent scenarios in which these networks will be deployed.\nReferences\n[1]\nBluetooth Special Interest Group (http://www.bluetooth.com/).\n[2]\nThe Wireless Personal Area Network Working Group, IEEE 802.15 (http://\nwww.ieee802.org/15/).\n" }, { "page_number": 388, "text": "Security in Wireless PAN Mesh Networks\n■\n377\n[3]\nJ. Haartsen, M. Naghshineh, J. Inouye, O. Joeressen, and W. Allen, Blue-\ntooth: Visions, goals and architecture, ACM SIGMOBILE Mobile Computing\nand Communications Review, Volume 2, Issue 4, 1998, pp. 38–45.\n[4]\nBluetooth Specification (https://www.bluetooth.org/spec/).\n[5]\nThe Wi-Fi Alliance (http://www.wi-fi.org/).\n[6]\nThe ZigBee Alliance (http://www.zigbee.org/).\n[7]\nIEEE 802.15.4-2003 Standard, Wireless Medium Access Control and Phys-\nical Layer Specifications for Low-Rate Wireless Personal Area Networks,\n2003.\n[8]\nJ. Daemen and V. Rijmen, The design of Rijndael—AES: The Advanced\nEncryption Standard, Springer-Verlag, 2002.\n[9]\nS. Seys, Cryptographic Algorithms and Protocols for Security and Privacy\nin Wireless Ad Hoc Networks, Ph.D. thesis, Katholieke Universiteit Leuven,\n2006.\n[10]\nA. Menezes, P. Van Oorschot, and S. Vanstone, Handbook of applied cryp-\ntography, CRC Press, 1996.\n[11]\nC. Gehrmann, J. Persson, and B. Smeets, Bluetooth security, Artech House,\n2004.\n[12]\nE. Filiol, Zero-knowledge-like Proof of Cryptanalysis of Bluetooth Encryp-\ntion, 2006.\n[13]\nB. Smeets and W. Chambers, Windmill generators—A generalization and an\nobservation of how many there are, Advances in Cryptology EUROCRYPT\n1988, Lecture Notes in Computer Science, Vol. 330, Springer-Verlag, 1988,\npp. 325–330.\n[14]\nB. Smeets and W. Chambers, Windmill PN-sequence generators, Computers\nand Digital Techniques, Volume 136, Issue 5, 1989, pp. 401–404.\n[15]\nH. Lipmaa, On differential properties of Pseudo-Hadamard Transform and\nrelated mappings, Progress in Cryptology, INDOCRYPT 2002, Lecture Notes\nin Computer Science 2551, Springer-Verlag, 2002, pp. 15–18.\n[16]\nY. Shaked and A. Wool, Cracking the Bluetooth PIN, 3rd International Con-\nference on Mobile Systems, Applications, and Services (MobiSys ’05), 2005,\npp. 39–50.\n[17]\nJ.L. Massey, G.H. Khachatrian, and M.K. Kuregian, SAFER+, Cylink Corpo-\nration’s Submission for the Advanced Encryption Standard, 1998.\n[18]\nJ.L. Massey, On the Optimality of SAFER+ Diffusion, Proceedings of the\n2nd Advanced Encryption Standard Candidate Conf (AES2), 1999.\n[19]\nNESSIE Project, New European Schemes for Signatures, Integrity, and En-\ncryption (http://www.cryptonessie.org/).\n[20]\nG. Lamm, G. Falauto, J. Estrada, and J. Gadiyaram, Security Attacks against\nBluetooth Wireless Networks, Second Annual IEEE Workshop on Informa-\ntion Assurance and Security, 2001, pp. 265–272.\n[21]\nD. Singel´ee and B. Preneel, Review of the Bluetooth security architecture,\nInformation Security Bulletin, Volume 11, Issue 2, 2006, pp. 45–53.\n[22]\nM. Jakobsson and S. Wetzel, Security Weaknesses in Bluetooth, Cryptogra-\npher’s Track at the RSA Conference (CT–RSA ’01), Lecture Notes in Com-\nputer Science 2020, Springer-Verlag, 2001, pp. 176–191.\n[23]\nDEF CON, Computer Underground Hackers Convention (http://www.\ndefcon.org).\n" }, { "page_number": 389, "text": "378\n■\nSecurity in Wireless Mesh Networks\n[24]\nH. Cheung, The Bluesniper Rifle, 2004.\n[25]\nA. Hodjat and I. Verbauwhede, The Energy Cost of Secrets in Ad-Hoc\nNetworks, IEEE Workshop on Wireless Communications and Networking\n(CAS ’02), 2002.\n[26]\nC. Candolin, Security Issues for Wearable Computing and Bluetooth Tech-\nnology, 2000.\n[27]\nC. De Canni`ere, T. Johansson, and B. Preneel, Cryptanalysis of the Blue-\ntooth Stream Cipher, COSIC internal report, Department of Electrical Engi-\nneering, Katholieke Universiteit Leuven, 2001.\n[28]\nN. Courtois and W. Meier, Algebraic Attacks on Stream Ciphers with Linear\nFeedback, Advances in Cryptology—EUROCRYPT 2003, Lecture Notes in\nComputer Science 2656, Springer-Verlag, 2003, pp. 345–359.\n[29]\nS. Fluhrer and S. Lucks, Analysis of the E0 Encryption System, 8th Annual\nInternational Workshop of Selected Areas in Cryptography (SAC 2001), Lec-\nture Notes in Computer Science 2259, Springer-Verlag, 2001, pp. 38–48.\n[30]\nJ. Golic, V. Bagini, and G. Morgari, Linear Cryptanalysis of Bluetooth\nStream Cipher, Advances in Cryptology—EUROCRYPT 2002, Lecture Notes\nin Computer Science 2332, Springer-Verlag, 2002, pp. 238–255.\n[31]\nM. Hermelin, and K. Nyberg, Correlation Properties of the Bluetooth Com-\nbiner Generator, 2nd International Conference on Information Security and\nCryptology (ICISC ’99), Lecture Notes in Computer Science 1787, Springer-\nVerlag, 1999, pp. 17–29.\n[32]\nF. Armknecht, J. Lano, and B. Preneel, Extending the Resynchronization At-\ntack, 11th Annual International Workshop of Selected Areas in Cryptogra-\nphy (SAC 2004), Lecture Notes in Computer Science 3357, Springer-Verlag,\n2004, pp. 19–38.\n[33]\nY. Lu, W. Meier, and S. Vaudenay, The Conditional Correlation Attack:\nA Practical Attack on Bluetooth Encryption, Advances in Cryptology —\nCRYPTO 2005, Lecture Notes in Computer Science 3621, Springer-Verlag,\n2005, pp. 97–117.\n[34]\nBluejacking (http://www.bluejackq.com/).\n[35]\nA. Laurie and B. Laurie, Serious Flaws in Bluetooth Security Lead to Dis-\nclosure of Personal Data, 2003.\n[36]\nD. Singel´ee and B. Preneel, The Wireless Application Protocol (WAP),\nInternational Journal of Network Security, Volume 1, Issue 3, 2005,\npp. 161–165.\n[37]\nJ. Kelsey, B. Schneier, and D. Wagner, Key Schedule Weaknesses in SAFER+,\n2nd Advanced Encryption Standard Candidate Conference, 1999, pp. 155–\n167.\n[38]\nD. Balfanz, D. Smetters, P. Stewart, and H. Wong, Talking to Strangers: Au-\nthentication in Ad hoc Wireless Networks, Network and Distributed System\nSecurity Symposium (NDSS 2002), The Internet Society, 2002.\n[39]\nJ. H. Hoepman, The Ephemeral Pairing Problem, Financial Cryptogra-\nphy, Lecture Notes in Computer Science 3110, Springer-Verlag, 2004,\npp. 212–226.\n" }, { "page_number": 390, "text": "Security in Wireless PAN Mesh Networks\n■\n379\n[40]\nJ. H. Hoepman, Ephemeral Pairing on Anonymous Networks, 2nd Inter-\nnational Conference on Security in Pervasive Computing (SPC 05), Lecture\nNotes in Computer Science 3450, Springer-Verlag, 2005, pp. 101–116.\n[41]\nD. Singel´ee and B. Preneel, Improved Pairing Protocol for Bluetooth, in\nProceedings of the 5th International Conference on Ad-Hoc Networks and\nWireless (ADHOC-NOW 2006), Lecture Notes in Computer Science 4104,\nT. Kunz, and S. S. Ravi (Eds.), Springer-Verlag, 2006, pp. 252–265.\n[42]\nF. Perez, Security in Current Commercial Wireless Networks: A Survey, 2006,\nhttp://www.hig.no/imt/file.php?id=1098/.\n[43]\nN. Sastry and D. Wagner, Security Considerations for IEEE 802.15.4 Net-\nworks, ACM Workshop on Wireless Security (WISE 04), 2004, pp. 32–42.\n[44]\nF. Stajano and R. Anderson, The Resurrecting Duckling: Security Issues in\nAd Hoc Wireless Networks, 7th International Workshop on Security Pro-\ntocols, Lecture Notes in Computer Science 1796, Springer-Verlag, 1999,\npp. 172–182.\n" }, { "page_number": 391, "text": "" }, { "page_number": 392, "text": "Chapter 12\nSecurity in Wireless LAN\nMesh Networks\nNancy-Cam Winget and Shah Rahman\nContents\n12.1\nIntroduction ........................................................ 382\n12.2\nWLAN Mesh Primer ................................................ 383\n12.3\nSecurity in WLAN Mesh Networks ................................. 385\n12.3.1\nWLAN Security Background .............................. 385\n12.3.2\nWLAN Mesh Security Primer .............................. 386\n12.4\nPossible Attacks on WLAN Mesh Networks ....................... 387\n12.4.1\nTypes of Attacks........................................... 387\n12.4.2\nAttacks on the Keys ....................................... 389\n12.4.3\nAttacks without Requiring Knowledge\nof the Secret Keys ......................................... 390\n12.5\nAttacks on WLAN Mesh Protocols ................................. 392\n12.5.1\nApproaches against Attacks on WLAN Mesh Protocols .. 392\n12.5.2\nAdvanced Attacks on WLAN Mesh Protocols ............ 393\n12.6\nOther Security Issues in WLAN Mesh Networks .................. 394\n12.6.1\nMesh Node Hijacking ..................................... 394\n12.6.2\nThreats from Bridged Networks .......................... 395\n12.6.3\nUnfairness from Greedy Nodes ........................... 395\n12.6.4\nNo Real Mutual Authorization ............................ 396\n12.6.5\nSupplicant–Authenticator Dilemma ....................... 396\n12.6.6\nAuthentication Server Location ........................... 396\n12.6.7\nManagement Frame Security .............................. 397\n12.7\nWLAN Mesh Security Requirements ............................... 398\n381\n" }, { "page_number": 393, "text": "382\n■\nSecurity in Wireless Mesh Networks\n12.8\nSecurity in IEEE 802.11s WLAN Mesh ............................. 400\n12.8.1\nThe Original IEEE 802.11s Proposal ...................... 400\n12.8.1.1 Overview ........................................ 400\n12.8.1.2 Security Framework ............................. 400\n12.8.2\nCurrent IEEE 802.11s Security Proposals ................. 401\n12.8.2.1 Proposal from Intel Corporation ................ 402\n12.8.2.2 Proposal from Tropos Networks\nand Earthlink .................................... 404\n12.9\nDiscussion and Conclusion ........................................ 405\nReferences................................................................. 406\nA technology that is sure to affect our lives significantly over the next\nfew years is wireless mesh networking. Wireless mesh as a technology\nhas been around almost as long as wireless LANs, but has only recently\nbecome more popular. As the popularity of wireless mesh networks grows,\nend users are demanding higher bandwidth, greater coverage, improved\nreliability, and robust security. The industry has come together at various\nIEEE 802 work groups to standardize wireless mesh networks with the right\ningredients and the right framework. Security is one of the cornerstones of\nmaking the disruption which is believed to be a reality with WLAN mesh\nnetworks. The WLAN mesh networking task group at IEEE codenamed TGs\nhas reached the first-draft specification stage, where security specification is\nnow essential. Security aspects of WLAN mesh networks entail a vast array\nof features and requirements to ensure that robust security is achieved at\nevery link of the mesh network. The roadmap for TGs is to develop a\nfull, official Extended Service Set or ESS mesh standard including mesh\ntransport security (versus end-to-end security) specifications targeted to\ncomplete around 2009 [1].\n12.1\nIntroduction\nWireless mesh networks have drawn a lot of attention in various market\nsegments, including home and small business networks, medium and large\nenterprise networks, public safety, emergency and first-responder networks,\nservice providers and wireless broadband networks, municipal and pub-\nlic access networks, and military and tactical networks. One of the core\ncomponents in making WLAN mesh networks successful and an enabler\ninto all these different markets is security. A core challenge in securing the\nWLAN mesh network is the large number of communication links over the\nair; as each mesh device is mobile and deployed outdoors, each mesh link\npresents an exposure and vulnerability into the mesh network.\nOriginal mesh architectures emerged from mobile ad hoc networks\n(MANETs) for military networks. The IETF MANET Work Group has been\n" }, { "page_number": 394, "text": "Security in Wireless LAN Mesh Networks\n■\n383\ndeveloping various MANET protocols for almost a decade [2–5]. MANETs\nwere envisioned to be military and tactical networks where peer nodes\ncould either come with or gain mutual trust between them. Mesh networks\nare different from MANETs in that there is more infrastructure commu-\nnication rather than direct, peer-to-peer communication with mesh net-\nworks becoming a popular deployment in public spaces. Especially in the\nmetropolitan space, existing IEEE networks’ security standards 802.1X [6]\nand 802.11i-2007 [29] based security mechanisms lack the specificity for\nsecuring the WLAN mesh network. Even though many vendors are using\nstrong 128-bit encryption to relay client and infrastructure traffic over the\nair, as previous wireless LAN attacks have shown, a cunning hacker may\nnot necessarily need to crack the key to get user information or damage the\nnetwork. Security researcher Shawn Merdinger says that municipal metro\ndeployments are going to be “a very serious security challenge to many\npeople” [8].\nThe rest of the chapter walks through the links and definitions in WLAN\nmesh networks from the security perspective; challenges and possible at-\ntacks in WLAN mesh networks; mesh client security; mesh infrastructure se-\ncurity; authentication, authorization, and access control; confidentiality and\nprivacy in mesh networks; and key management in WLAN mesh networks.\n12.2\nWLAN Mesh Primer\nIt is important to carefully define WLAN mesh components and segments\nfor examining the security implications on the overall mesh network. From\na security perspective, there are two major components of a mesh network:\n1.\nA wired or bridged segment: The network attached to a mesh net-\nwork and that operates over the wire, e.g., Ethernet or fiber. One\nor more of these segments may be attached to a mesh network.\n2.\nA wireless or mesh segment: The all-wireless network that may or\nmay not be attached to a wired or bridged segment. The transport\nmedia of this segment is IEEE 802.11 for WLAN mesh networks.\nThis segment is commonly referred to as a mesh network.\nWired and bridged segments of the network are generally considered\noutside the scope of a WLAN mesh network. However, they may impact se-\ncurity in a mesh network by launching attacks or injecting carefully crafted\nframes into it. Hence, it is important to secure the entry points from these\nsegments into a mesh network.\nThe mesh segment of the network requires careful security considera-\ntions as it is exposed to attackers as frames are transmitted over the air.\nThere are two major sub-components of this segment:\n" }, { "page_number": 395, "text": "384\n■\nSecurity in Wireless Mesh Networks\n1.\nMesh backhaul: A mesh backhaul consists only of mesh nodes and\nmesh links. This is an all-wireless, multi-hop network helping WLAN\nclient traffic to traverse over 802.11 links to and from a wired entry\npoint or other WLAN clients in mesh.\n2.\nMesh access: A mesh access consists of mesh nodes co-located with\nWLAN access points and WLAN clients. This single-hop network\nallows end users to connect to a mesh network.\nA mesh node is a physical or logical entity in a mesh network partic-\nipating in formation of a mesh. TGs define mesh nodes as either a mesh\npoint (MP, capable of forming links between mesh nodes only) or a mesh\naccess point (MAP, capable of forming links between mesh nodes as well\nas links between mesh nodes and WLAN clients). There is a special mesh\nnode, which interfaces a mesh network to wired or non-WLAN bridged\nnetworks, called mesh portal or MPP. Common mesh node architectures\ninclude:\n■\nSingle-radio node: A mesh node consisting a single IEEE 802.11b/g\nor 802.11a radio. This node commonly is an MAP allowing user\naccess on the same radio where mesh backhaul links are formed.\nAn MP with a single-radio allows only mesh backhaul links over its\nradio.\n■\nDual-radio node: A mesh node consisting of two IEEE 802.11b/g\nor 802.11a radios (in any combination), one dedicated for forming\nmesh backhaul links, the other dedicated for allowing user access.\nThis architecture is common today where lower-capacity and lower-\ncost radio (such as 802.11b) is used for client access and higher-\ncapacity radio (such as 802.11a) is used for mesh backhaul.\n■\nMulti-radio node: A mesh node consisting of multiple IEEE 802.11b/g\nor 802.11a radios (in any combination). Multiple radios may be used\nfor allowing user access and multiple radios may be used for mesh\nbackhaul. Typically, mesh backhaul forming on different radios\ndedicates one radio for frame transmission and another for frame\nreception. Another common division of labor occurs for separating\nupstream and downstream traffic of mesh backhaul to and from an\nMPP.\nA mesh link is a logical 802.11 WLAN link between two MPs or MAPs.\nAn access link is a logical 802.11 WLAN link between an MAP and a WLAN\nclient. A mesh network consists of both types of links, whereas a mesh\nbackhaul consists only of mesh links. Typically, access links are simple\nradio-links set up and operated according to IEEE 802.11 standards. The\nmesh links are more complicated and two mesh nodes can have connec-\ntions over multiple radios. Such links are common in mesh networks where\n" }, { "page_number": 396, "text": "Security in Wireless LAN Mesh Networks\n■\n385\nWireless/mesh segment\nMesh path\nMesh node\nMesh link\nAccess\nlink\nMesh access\nMesh backhaul\nWired/bridged segment\nBridged link\nFigure 12.1\nLinks, nodes, and segments for WLAN mesh security.\nmulti-radio mesh nodes are deployed. Figure 12.1 shows how all different\ncomponents and segments come together in a WLAN mesh network.\n12.3\nSecurity in WLAN Mesh Networks\nBecause WLAN mesh networks are based on original WLAN networks, we\nfirst look at WLAN security protocol standards and how they are deployed.\nWe then examine how and where these protocols are not sufficient for\nWLAN mesh security.\n12.3.1\nWLAN Security Background\nThe first IEEE 802.11 standard included a weak security protocol called WEP\n(Wired Equivalent Privacy), which failed to provide the goal of wired equiv-\nalence [26,28]. These flaws and the adoption of NIST-approved ciphers\nwere addressed by the ratification of the IEEE 802.11i [7] amendment in\n2004 and its inclusion in the base IEEE 802.11-2007 specification [29]. Prior\nto the ratification, the industry also embraced an early version of 802.11i to\nprovide a migration path to 802.11i. The wireless alliance WiFi embraced\nthis migration path and referred to it as WPA (Wireless Protected Access).\nThe major weaknesses of WEP include:\n1.\nLack of mutual authentication\n2.\nNo access control\n3.\nNo replay prevention\n" }, { "page_number": 397, "text": "386\n■\nSecurity in Wireless Mesh Networks\n4.\nNo message modification detection\n5.\nCompromised message privacy due to IV reuse, RC4 weak keys,\nand possibility of direct key attacks\nFor details on these weaknesses, look at Chapter 6 of [9], which pro-\nvides an in-depth analysis of them; alternately [28] provides a comprehen-\nsive summary. IEEE 802.11i defines a new type of wireless network called\nan RSN (Robust Secure Network). To allay industry concerns for already-\ndeployed systems, the WiFi alliance took a subset of 802.11i and created\nWPA while allowing the IEEE 802.11 standards body to focus on a sound,\nlonger-term solution. As WPA is a subset of 802.11i, they both provide a\nframework referred by 802.11i as RSN. The framework allows for the ne-\ngotiation of authentication, key management, and cipher suites used to\nultimately protect the 802.11 link. While the RSN framework enables pro-\nprietary mechanisms to coexist, it defines the following components:\n1.\nAuthentication and key management: The mandatory-to-implement\nmechanism is based on IEEE 802.1X to enable Extensible Authenti-\ncation Protocol (EAP) methods to be used for authentication. Simi-\nlarly, IEEE 802.1X is used to employ a key management mechanism\nto allow the client and access point to mutually derive the keying\nmaterial needed to protect the 802.11 link and subsequent 802.1X\nkey management functions. Optionally, an RSN also enables the use\nof pre-shared keys as a replacement to EAP for those systems that\ndo not have the back-end infrastructure for identity management.\n2.\nCipher suite: The mandatory-to-implement cipher suite is based on\nAES-CCM and Temporal Key Integrity Protocol (TKIP) is provided to\nallow already-deployed systems to allay the vulnerabilities of WEP.\n12.3.2\nWLAN Mesh Security Primer\nIn the past, security architectures were often developed based on the as-\nsumption that the core parts of the network were not physically accessible\nto an enemy. Attacks were only expected to be launched in well-defined\nplaces such as connections to the public Internet. Firewalls and intrusion\ndetection systems were deemed sufficient to keep valuable electronic assets\nin a corporation or personal data from being stolen, exposed, or compro-\nmised. WLAN networks break this conventional assumption in network\nsecurity. Because data now passes over radio waves, ready and easy ac-\ncess to data becomes trivial. Original WLAN technology was targeted for\nindoor LAN networks, keeping the sphere of exposure somewhat limited,\nalthough unpredictable radio waves do propagate outside the buildings.\nWar-driving and sniffing near buildings may allow an attacker to see much\nof the data traveling inside the buildings, too. Sniffing is defined as simply\n" }, { "page_number": 398, "text": "Security in Wireless LAN Mesh Networks\n■\n387\nusing a software and WLAN radio card to read and store all frames flowing\nover a WLAN channel.\nOutdoor WLAN networks exacerbate security exposure by deliberately\ntransporting data over radio waves through open air in metropolitan and\nrural areas; that is, exposing the physical access points in the open public.\nIn other words, now an attacker does not need to drive closer to the build-\nings anymore. Anyone can see those radio waves and its data at will from\nanywhere in a city or rural area wherever those radio waves traverse or ac-\ncess the exposed access points from the street. Whether indoor or outdoor,\nmesh networks may take the strategy of re-using 802.11i for mesh access.\nBut this leaves mesh backhaul not secured and there is no standard mech-\nanism for securing mesh backhaul today. There is also the need to secure\nperipheral devices attached to the wired interfaces of mesh nodes. Finally,\nmobility of WLAN clients and mesh nodes makes mesh security a great\nchallenge in defining an interoperable standard. Vendors are currently off-\nering proprietary mechanisms for backhaul and bridge security restricting\nsingle vendor mesh deployments presenting a hurdle toward widespread\nadoption of secure WLAN mesh networks.\n12.4\nPossible Attacks on WLAN Mesh Networks\nThis section examines possible attacks and threat models in WLAN mesh\nnetworks. Many of these attacks are similar to that of attacks in WLAN net-\nworks. Attacks on wireless networks can be classified into five broad cat-\negories: eavesdropping, forgery, masquerading, man-in-the-middle (MIM),\nand denial of service (DoS). The first category of attack is also known as\npassive, the other three are known as active attacks. Some in-depth attack\nscenarios and analysis of those scenarios would be useful in understanding\nand deriving the mechanisms needed to prevent these attacks and protect\nthe network against them.\n12.4.1\nTypes of Attacks\nEavesdropping is accessing information without detection of either the data\noriginator or the intended receiver. More importantly, it is information to\nwhich the attacker does not have legal access. Such information may in-\nclude confidential company data, personal financial and medical informa-\ntion, etc. An attacker may sniff data over 802.11 channels in either a mesh\naccess or backhaul network. Especially in a wireless medium, this form of\nvulnerability enables an attacker to gain information without detection from\nany of the communicating parties and is typically referred to as a passive\nattack.\n" }, { "page_number": 399, "text": "388\n■\nSecurity in Wireless Mesh Networks\nForgery is the ability to change any content of a frame without detec-\ntion. Such modification can cause a frame to be redirected to a different\nsource or, more damaging, change the original information to the intended\nreceiver. Although protection from eavesdropping can help, equally dam-\naging is the ability for an attacker, for example, to forge a stock transaction\nfrom a buy to a sell order.\nMasquerading (sometimes referred to as spoofing) occurs when an at-\ntacking network device impersonates a valid device. Depending on whether\na device is accessing a mesh node using its MAC or IP address, an atta-\ncker may either use IP address spoofing or MAC address spoofing. Noto-\nrious attacks, such as evil twin attacks, can potentially allow hackers to\nsteal personal information such as credit cards or any personal identity\ninformation.\nMan-in-the-middle can be another form of a forgerer, a masquerader,\nand even an eavesdropper. An MIM attacker interjects communication by\npretending to be the network to the client and the client to the network.\nBy interjecting the communication, neither the client nor the network may\nbe aware that the MIM can now gain identity information from the client\nand potentially launch other attacks against the network.\nDoS attacks work with the principle of causing damage to the target\ndevice or the overall network itself. In wireless, DoS attackers can simply\njam the radio frequency. In general though, DoS attackers often target some\nnodes in a network and overwhelm them with traffic, eventually causing\nthem to reboot or melt down. ICMP flood or Ping of Death are examples of\nclassic DoS attacks, which the Internet experienced in the 1990s. A variation\nof DoS, distributed DoS (DDoS) attacks are more effective where attackers\nlaunch DoS traffic from several zombie computers from different locations.\nWhile DoS and DDoS attacks are easy to mount in WLAN networks and in\nmesh networks, they are almost impossible to prevent. Because most WLAN\nmesh networks run in the unlicensed 2.4 and 5 GHz bands, hackers may\nnot even need to use WiFi to conduct DoS attacks against these networks.\nEspecially in municipal networks where free WLAN infrastructures are now\nin place outdoors, more and more esoteric attacks will come into play.\nFor example, widespread Bluetooth attacks and Bluetooth spamming are\nreal possibilities with WLAN mesh networks combined with small PCs like\nGumStix with Bluetooth.\nAlthough it may be more challenging to ward off all DoS attacks, WLAN\nsecurity must address protection from eavesdropping, forgery, masquerad-\ning, MIM, and, where feasible, DoS attacks.\nIn further providing security mechanisms, attacks on such protective\nmeans must also be addressed. As most systems employ the use of a known\nsecret referred to as a key, considerations for the threats against the very\ncryptographic tools used to provide security also merit description. These\nattacks are categorized as:\n" }, { "page_number": 400, "text": "Security in Wireless LAN Mesh Networks\n■\n389\n1.\nAttacks to recover the secret key\n2.\nAttacks with limited or no knowledge of the secret keys\n12.4.2\nAttacks on the Keys\nThe challenge in any cryptographic tool employing shared keys is to en-\nsure that these keys are strong enough and not susceptible to its recovery.\nBecause the shared key is used to gain access to the network or to protect\nthe communication with the network, it is critical that it be very difficult\nto recover these keys; otherwise, knowledge of the key often represents a\nfull breach in security [9]. In real-world use, these keys may oftentimes be\nrequired to be manually entered, especially when used as a means to iden-\ntify a user. In this scenario, these keys are often referred to as passwords\nas people usually choose something that can be easily remembered.\nAs passwords tend to be derived from a language source of finite vocab-\nulary, tools based on dictionary attacks can be readily employed to break\nsuch keys. Other, more complicated attacks can analyze the actual func-\ntions used to derive the keys, or how the keys are actually employed to\nrecover the actual key. The original (flawed) IEEE 802.11 security protocol\nWEP constructed its protocol in such a way that it was easy to recover the\nkey [26]. Though such attacks require some data sampling, this require-\nment is trivialized in WLAN mesh networks as the data is easily obtained\nby capturing the signals over the air.\nAttacks on keys are beneficial and worth pursuing especially if the\nstrength (e.g., entropy) of a key is known to be weak. Some techniques of\nattacking on the keys include:\n1.\nBrute-force method: An attacker tries every possible key until he\nfinds a match. Guessing passwords is an example of such attacks.\nThe time taken for a brute-force attack depends on key entropy.\nHence, making the key-size longer does not always solve the prob-\nlem (it only takes longer to break the key).\n2.\nDictionary method: An attacker uses a dictionary, or database, con-\ntaining all the likely passwords/keys. Sometimes known as an off-\nline attack, an adversary can take known matching ciphertext and\nplaintext and run a computer and a dictionary loaded to find the\nkeys, which produces the ciphertext from the given plaintext. IEEE\n802.11i key derivation makes keys dynamic and usable only for a\nsingle session to reduce the chance of such attacks. WLAN mesh\nnetworks should not be susceptible to dictionary attacks if similar\nsession key derivation mechanisms are used.\n3.\nAlgorithmic method: Adversaries also have the actual cryptographic\nalgorithms and frame constructions from which they can analyze,\nas was shown by Fluhrer et al. [26] to demonstrate weaknesses in\n" }, { "page_number": 401, "text": "390\n■\nSecurity in Wireless Mesh Networks\nthe algorithm and aid in key recovery. There are also optimizations\non dictionary attacks that enable smaller or more exhaustive dic-\ntionaries and variations to be used by trading memory and space\n[27].\nAs many tools for cracking WEP are now readily available and with\nthe wider adoption of IEEE 802.11i, WLAN mesh networks must not con-\nsider WEP for either infrastructure, access, or ad hoc security. With the\nlevel of exposure in metro and outdoor areas, cracking WEP would be\ntrivial for attackers of WLAN mesh networks. Note that some WLAN client\ndevices like cameras (e.g., D-Link IP Camera and Linksys Wireless-G\nInternet Video Camera) and video game consoles (e.g., Linksys Wireless-B\nGame Adapter and Xbox 360 Wireless Networking Adapter) continue to im-\nplement WEP-based encryption only. These devices should not be allowed\nto connect to WLAN mesh networks. Fortunately, many client devices like\nCannon SD430 Powershot Camera now support advanced 802.11 encryp-\ntion, e.g., AES-CCMP which is part of the IEEE 802.11i standard. Over time,\nall WLAN client devices should migrate to these more-robust encryption\nmethods.\n12.4.3\nAttacks without Requiring Knowledge\nof the Secret Keys\nIronically, all five types of attacks described earlier in the section can be\nconducted without or with limited knowledge of these keys. Even en-\ncrypted traffic can reveal information such as how, when, and by which\ndevices the network is being used. Another example is that of manage-\nment frames, especially beacons and probe responses as they are never\nencrypted and where an attacker can readily learn the SSID being broad-\ncast by mesh node or manufacturer, model, and other device information of\nthe node encoded in 802.11 information elements. The attacker may exploit\nany known vulnerabilities in that particular model hardware or software.\nFor example, there may be open-source security software libraries (e.g.,\nopenSSH [10] and openSSL [11]) in cheaper mesh nodes and the attacker\nmay have the knowledge of public-domain vulnerabilities which can be\neasily exploited. The attacker can also perform sophisticated traffic analy-\nsis by studying message externals, e.g., frequency of communication, size\nof payload, traffic load on a device, etc. Finding a correlation of TCP ac-\nknowledgment frames or DHCP discover messages, which are of fixed\nlength and might occur at regular intervals, provides a wealth of informa-\ntion to the attacker. Typically, such information is useful in conjunction\nwith other techniques, such as modification.\n" }, { "page_number": 402, "text": "Security in Wireless LAN Mesh Networks\n■\n391\nIn a secure WLAN where packets are encrypted, forgery and MIM attacks\nare difficult to mount against networks because the attacker must intercept\ntransmission from either end (AP or client) and relay it without giving any\nclue to the receiver about the compromise. This is done in turn for both\nends creating a relay or repeater node in between the AP and client. In a\nWLAN mesh network, an MIM attack can be launched between MP links\nas well. MIM between mesh nodes would be more damaging compared to\na compromised AP and client link because now all backhaul traffic over\nthe compromised mesh link is affected. A carefully crafted attack may get\nan MP in thinking of a rogue device to be valid and relay traffic to and\nfrom it. In this attack, the adversary can either direct traffic to its intended\ndestination or mess with the data. Both strategies impact the services of a\nmesh network, more so if the MIM is in between mesh nodes.\nAnother threat emerges from the ability to replay messages either to\nthe network or to the endpoint device. The attack could be maliciously or\nfraudulently repeated by either the originator or an MIM.\nDoS and DDoS attacks do not require knowledge of the shared secret,\nespecially in WLANs. An attacker or a group of attackers launch these\nattacks simply to bring down a network or its services. WLAN mesh net-\nworks are particularly susceptible to these attacks and present a great chal-\nlenge. A special type of DoS attack known as RF jamming against WLAN\nnetworks is very difficult to detect and prevent. More damaging is the cur-\nrent lack of protection for 802.11 management frames. Two such frames,\nDisassociation and Deauthentication, permit using the broadcast MAC ad-\ndress as the target and are easy means to disrupt WLAN service to all\nconnected clients of the victim access point. These frames may also be di-\nrected to a specific station, denying service to targeted victims. Similarly, an\nattacker may observe the victim station’s MAC address and send an Asso-\nciation Request to a different AP on the same wired LAN. This association\nrequest is accepted as if the station is roaming and the wired network now\nforwards all traffic to the attacker. In some networks, the victim station\nmay be disconnected from the AP it was attached to and, depending on\nthe security method negotiated, the adversary may not be required to re-\nauthenticate with the new AP. Yet another example is where an adversary\nuses a station simulator tool, such as the Veriwave WLAN Simulator, and\ncongests an AP with bogus stations exhausting its available resources over\nthe air, eventually causing the victim AP to stop accepting new clients or,\nin some implementations, to reboot. Clever attackers may continually keep\nloading bogus stations on the AP, completely taking it out of service.\nAll these classic DoS/DDoS attacks are more easily applicable to WLAN\nmesh networks because adversaries now have visibility into client traffic\nstreams from anywhere in a mesh deployed area. Many other possible atta-\ncks on WLAN and ad hoc networks without keys are described in [21–23].\n" }, { "page_number": 403, "text": "392\n■\nSecurity in Wireless Mesh Networks\n12.5\nAttacks on WLAN Mesh Protocols\nWLAN mesh networks face another array of security challenges that emerge\nfrom its multi-hop nature. The default routing protocol in TGs is Hybrid\nWireless Mesh Protocol or HWMP, which provides the ability for a mesh\nnode to learn routes to another mesh node using a broadcast route dis-\ncovery mechanism. Broadcast-based route discovery mechanisms are tradi-\ntionally susceptible to DoS attacks as they use exhaustive re-broadcasting\nmethods. An attacker may snoop frames over a WLAN mesh backhaul and\nlearn about MAC addresses or various mesh nodes in the network. Because\nHWMP is based on the IETF’s AODV [12], an open-source AODV software\nstack can be used to continually generate route request (RREQ) frames\nkeeping all mesh nodes in the network busy re-broadcasting those. This\nmay cause one or more mesh nodes to melt down, reboot, or stop servicing\nthe network.\nOther attacks possible on an unprotected RREQ include:\n■\nRoute disruption by changing message type, destination address,\nsource address, or originator address\n■\nRoute invasion by increasing RREQ-ID, originator sequence number,\nor destination sequence number by at least one\nAttacks on route replies (RREP) are possible when the attacking node\ndrops all routing frames, causing the routes to take longer and sub-optimal\npaths. Often an attacking device positioned in between valid devices may\ncut off some routes all together. MIM attacks are possible if particular route\ndestinations can be lured to an adversary’s device followed by a detour\nsomewhere over the Internet. The attacker may do so by sending fake\nRREPs with a large enough destination sequence number or short hop\ncount.\nAttacks on route errors (RERR) are not as severe because the result is\nroute disruption. Yet, generating bogus RERRs can cause many nodes to\nattempt to repair processing and re-discover valid routes. Another point\nto note is that most fields of RRER, RREP, and RERR, e.g., ID, Hop Count,\nMetric, Sequence Number, etc., are vulnerable to modification and forgery.\nMost damaging is the vulnerability of an MAC address, as an adversary\ncan impersonate an MP by simply using its MAC address; an adversary can\nsimply form part of mesh forwarding paths and launch any attack from\nthere. Note that similar attacks are also possible against RA-OLSR, which is\nthe optional path selection protocol in IEEE 802.11s draft standard.\n12.5.1\nApproaches against Attacks on WLAN Mesh Protocols\nEven when mesh nodes are authenticated before joining a WLAN mesh\nnetwork, many aspects of a mesh are controlled via broadcast frames.\n" }, { "page_number": 404, "text": "Security in Wireless LAN Mesh Networks\n■\n393\nIn a broadcast environment, all parties can discern the information and\noften can affect other members of the group. An insider attack is a form\nby which an adversary may be able to join the mesh by exploiting weak-\nnesses in the mesh authentication mechanism and exploit the broadcast en-\nvironment to launch attacks. Broadcast protocols in the IEEE 802.11s draft\nstandard do not have any mechanism for protecting themselves from in-\nsider attacks. There are techniques for protecting HWMP by using methods\nsuch as authenticated broadcast of RREQ, authenticated unicast of RREP,\nand authenticated broadcast of RERR. On top of node-based authentication\nof routing nodes, individual message integrity and authenticity are also\nneeded to limit and prevent the attacks described earlier. SAODV [13] is a\nsecure version of the original AODV protocol, which combines these tech-\nniques and more (e.g., digital signature for static fields in headers and hash\nchains to protect Hop Count). While SAODV is appropriate for ad hoc net-\nworks, it comes with some costs for WLAN mesh networks. Even though\nhash chains are efficient for Hop Count authentication, a malicious node\ncan still choose not to increase it. Other drawbacks of SAODV include PKI\ninfrastructure usage and key distribution, too frequent signature computa-\ntions, and extra overhead for exchanging signatures, which can be up to\ntwo signatures per message, becomes computationally prohibitive. At the\ntime of this publication, IEEE 802.11 TGs is evaluating these techniques\nand may incorporate some subset of SAODV for securing the default path\nselection protocol, HWMP.\nARAN [14] and Ariadne [15] are two other published techniques for\nsecuring AODV, which can be adapted for securing HWMP.\n12.5.2\nAdvanced Attacks on WLAN Mesh Protocols\nIn addition to the attacks previously discussed in this chapter, attacks tar-\ngeted to peer-to-peer or mesh networks may also be applied to WLAN\nmesh networks. These attacks can be summarized as follows:\n■\nSybil attacks: An adversary presents itself as being multiple illegit-\nimate identities to the mesh network. Thus, given a single faulty\nentity, it can masquerade as many other entities and control a part\nof the network. This attack requires that each MP be provisioned\nwith strong authentication identification and authentication of the\ntraffic being routed within the mesh.\n■\nSinkhole attacks: An attacking node lures all traffic around it by\ninstalling an attractive node. Powerful transmitters and high-gain\nantennas may allow the device to emerge as high-quality routes.\nSinkhole attacks open doors for further ugly attacks and tamper-\ning with application data. Detection of sinkholes is difficult without\nhigher-layer protections such as asking for acknowledgments from\n" }, { "page_number": 405, "text": "394\n■\nSecurity in Wireless Mesh Networks\nthe final destinations for all messages (TCP and HTTP implement\nacknowledgments as part of the base protocols). Sinkhole devices\nare often referred to as honeypots.\n■\nBlack hole/gray hole attacks: An attacking node drops all frames it\nreceives (black) or drops selective frames it receives (gray). In black\nhole attacks, mesh nodes can protect themselves by requesting ex-\nplicit acknowledgment for routing protocol and application frames.\nGray hole attacks are more challenging to detect because the attack-\ning node appears as a valid forwarder. Higher-layer protocols end up\nsuffering from the dropped frames, which may degrade application\nquality (e.g., for UDP streams) or cause excessive retransmissions\nand shrinkage of data burst windows used by transport layer proto-\ncols, e.g., sliding window in TCP.\n■\nWormhole attacks: An attacker may leverage multiple attacking\nnodes and create low-latency and high-speed route tunnels be-\ntween them. This strategy will make attacker’s tunnel appear at-\ntractive over a multi-hop path and cause a wide area of nodes to\nattempt to use the tunnel. Black hole/gray hole/sinkhole attacks\nmight follow. Unfortunately, wormhole attacks are effective even if\nthe protocol/system provides authenticity and confidentiality.\nGiven the use of strong identification credentials, e.g., strong entropy\nkeys and unique identities, IEEE 802.11 TGs may be able to address some\nof the above attacks, but may still be susceptible to insider attacks.\n12.6\nOther Security Issues in WLAN Mesh Networks\nIn addition to the various WLAN and WLAN mesh attacks described in pre-\nvious sections and approaches in solving those, there are further security-\nrelated issues that exist in practical WLAN mesh networks:\n■\nMesh node hijacking\n■\nThreats from bridged networks\n■\nUnfairness from greedy nodes\n■\nNo real mutual authorization\n■\nSupplicant-authenticator dilemma\n■\nAuthentication server location\n■\nManagement frame security\n12.6.1\nMesh Node Hijacking\nIn a WLAN mesh network, route paths and topologies can be arbitrarily\nestablished independent of the path selection protocols: HWMP or OLSR.\nBecause there is no administrative boundary or domain enforced by these\n" }, { "page_number": 406, "text": "Security in Wireless LAN Mesh Networks\n■\n395\nprotocols, different ISP networks that can see each others’ mesh nodes may\nend up proliferating into each others’ network. A greedy network owner\nmay attempt to leverage other owners’ mesh nodes for forwarding its own\ntraffic. A hostile network owner may attempt to leverage neighbor owners’\nmesh nodes for forwarding its own traffic and take one step further that\nprotects its own mesh nodes by proprietary means. HWMP should con-\nsider defining administrative boundaries like routing protocols used in the\nInternet, e.g., Border Gateway Protocol (BGP) or Open Shortest Path First\n(OSPF).\n12.6.2\nThreats from Bridged Networks\nIn a WLAN mesh network, many nodes are equipped with Ethernet or fiber-\nwired interfaces. A greedy network owner may install a large wired LAN to\nits mesh node and connect to the network. Because there is no standard\nmethod of authenticating the devices connected to these interfaces of a\nmesh node, this poses a security challenge on these open ports. Unless\nthere is an authentication server (AS) in the mesh node, it will have to reach\nout to some remote AS inside or outside the WLAN mesh to authenticate\nthe devices connected to these interfaces. If there is a reachable AS, the\nnode may employ IEEE 802.1x [6] port control mechanisms on it. There\nare still open issues as to which devices should be authenticated and how\nmany, as there may be an entire switched or bridged LAN behind those\nwired interfaces.\nAnother threat from bridged networks occurs when there are two wired\nLANs connected to the same WLAN mesh network and they start using the\nmesh as a wireless bridged network. Because there is an inherent mismatch\nbetween wire speed of wired LANs and shared media in WLAN mesh, this\nmay seriously starve traffic in a WLAN mesh network or even simulate a\nDoS-attacked WLAN mesh.\n12.6.3\nUnfairness from Greedy Nodes\nAs mesh nodes may relay traffic for their own clients as well as for other\nmesh nodes, throughput obtained by them may significantly vary depend-\ning on their position in the network. This is particularly true for a hierarchi-\ncal mesh where most communication occurs to and from a limited number\nof MPPs. Usually, nodes further away from the portals suffer highly un-\nfair and degraded throughput. This implies degrading quality of service for\nthe clients farther away from MPPs. Currently, there is no solution to this\nproblem in the IEEE 802.11s draft standard.\nAn attacker with knowledge of a mesh hierarchy may exploit the fact\nand start installing greedy nodes anywhere in the hierarchy with a mission\nof further starving or completely blocking out access to nodes farther from\n" }, { "page_number": 407, "text": "396\n■\nSecurity in Wireless Mesh Networks\nMPPs. They may appear as hidden nodes to the suffering nodes, winning\n(or jamming) the channel and causing excessive collisions.\n12.6.4\nNo Real Mutual Authorization\nIn a WLAN mesh network, it is difficult to ascertain what service data for-\nwarding, service clients, etc., the nodes are authorized to. Even though\nserver-based policies can be used to provide proper authorization for a\nnew mesh node, there is no mechanism for the mesh node to authorize\nother members in the network or to learn of their peer authorizations. This\nmay result in a mesh node to join an alien network and become a slave.\n12.6.5\nSupplicant–Authenticator Dilemma\nThe EAP security mechanisms [16] are widespread not only in the IP-based\ndata communication world, but also in cellular and other parts of the wire-\nless communications world. EAP works based on a three-party model at-\ntempting to authenticate a node in a network (supplicant) via an already\nauthenticated node (authenticator) by an AS. If there is an AS present in\nthe network, whichever node has an active connection to the AS takes up\nthe role of authenticator and the other becomes a supplicant. In a mutual\nauthentication scenario, the roles would have to be swapped for the nodes\nto be fully and mutually authenticated using an EAP method. This scheme\nrequires implementing both supplicant and authenticator stacks in every\nnode, causing code and other resource bloats, such as system memory.\nOne alternative to avoid this problem is to use a fixed authenticator in\nthe network, e.g., a portal device, and let authenticated nodes pass through\nfor nodes which join the network. This method requires implementing only\nthe supplicant stack on mesh nodes while implementing authenticator stack\nat selective mesh nodes, such as a portal. Another alternative is to avoid the\nuse of EAP for authentication and use a peer-based mutual authentication\nmethod.\n12.6.6\nAuthentication Server Location\nAS location and setup is another open issue in WLAN mesh networks. An\nAS can be located inside or outside a WLAN mesh network. The location\nof the AS affects re-authentication unless there is optimization to avoid\ninvolving the AS in the re-authentication process. If the AS is located inside\nthe mesh, all mesh nodes must be aware of where it is. If the AS is outside\nthe mesh, only portals need to know where it is. The number of ASs and\norientation also affects WLAN mesh security. For example, a centralized\n" }, { "page_number": 408, "text": "Security in Wireless LAN Mesh Networks\n■\n397\nAS can be used for authentication, authorization, and access control (AAA)\nof all mesh nodes. Similarly, a distributed AS model can be used where\nmultiple ASs provide AAA services in mesh. TGs is not specifying any\nparticular AS deployment model for WLAN mesh networks.\n12.6.7\nManagement Frame Security\nThe final topic we examine in WLAN mesh security issues is the securing of\nmanagement frames as these frames are the foundation for many DoS at-\ntacks against early 802.11 WLAN networks. IEEE 802.11 has already formed\na Task Group W to address this need for the general 802.11 management\nframes. The objective for management frame security in a WLAN mesh\nis to assure authenticity, integrity, and privacy (where appropriate) of the\nmanagement frames sent and received among MPs on a link-by-link basis.\nThe IEEE 802.11i-based link level authentication model can be leveraged to\nsupport authentication, key distribution, and encryption for management\nframes. There is unlikely to be any separate management frame specific au-\nthentication and encryption architecture. Management frames should have\nthe same level of security and use the same mechanisms as data frames.\nWherever possible, the security mechanisms defined by the Task Group\n802.11w [19] will be utilized. WLAN mesh management frame protection is\nused for the following purposes in a WLAN mesh network:\n1.\nForgery protection\n2.\nConfidentiality protection\n3.\nCompatibility with 802.11i key hierarchy\n4.\nIncremental inclusion of new management frames\n5.\nProtection only after key establishment\n6.\nFragmentation support for management frames\nWhen considering security, the mesh management frames as well as\n802.11 standard [20] management frames can be classified in two broad\ncategories:\n1.\nThose sent prior to authentication\n2.\nThose sent once 802.11 link layer is secured\nThe management frames sent prior to authentication are Mesh Bea-\ncon, Probe Request/Response, 802.11 and 802.1X Authentication Request/\nResponse, Association Request/Response, and the 802.11i four-way hand-\nshake. When 802.1X EAP is used, the management frames used are not\nprotected at the link layer. The management frames sent and received\nafter authentication are Mesh Beacon, Reassociation Request/Response,\nATIM, Disassociation, Deauthentication, action management frames and\n" }, { "page_number": 409, "text": "398\n■\nSecurity in Wireless Mesh Networks\nmesh-specific management frames. All these frames should be secured us-\ning 802.11w and derivative techniques.\n12.7\nWLAN Mesh Security Requirements\nNow that typical security threats and attacks in WLAN mesh networks have\nbeen discussed and analyzed, WLAN mesh security requirements can be\nderived in a methodical manner. From a high-level perspective, they can\nbe first categorized into the following four broad categories:\n1.\nInfrastructure security: Data, control, and management traffic secu-\nrity that flows over the infrastructure mesh nodes and mesh links.\nThis is often termed “backhaul security.”\n2.\nNetwork access security: Data, control, and management traffic se-\ncurity that flows between a WLAN client and MAP.\n3.\nAd hoc security: Data, control, and management traffic security that\nflows between two WLAN clients over a multi-hop path in a mesh\nnetwork. In many cases, MAPs and clients may be mobile and sus-\nceptible to dynamic topology changes in mesh backhaul or network.\n4.\nApplication security: Security of the applications run by WLAN\nclients in a mesh network, such as VoIP, database, etc.\nAmong these security categories, ad hoc security is by far the most chal-\nlenging of all. Application security is typically not addressed within the net-\nwork stack and is implemented by the applications at network endpoints.\nWith respect to the other three categories, the WLAN security requirements\ncan be stated as follows:\n1.\nMesh node and client authentication: A mesh node should authenti-\ncate a requesting WLAN client before servicing it. The WLAN client\nshould also authenticate the mesh node to avoid joining rogue mesh\nnodes. This mutual authentication requirement is needed to prevent\nunauthorized network access from both mesh node and client per-\nspectives.\n2.\nMesh node and client key agreement: A mesh node and client\nshould undergo handshakes to establish a fresh shared key to en-\ncrypt, authenticate, and integrity protect all traffic flowing between\nthe mesh node and the client. This key must be a short-lived key that\nis freshly derived when the session is initiated and deleted once the\ncommunication between the mesh node and client is terminated.\n3.\nMesh node and mesh node authentication: A mesh node should au-\nthenticate another mesh node before forwarding traffic to and from\nit. The joining mesh node should also authenticate any other mesh\n" }, { "page_number": 410, "text": "Security in Wireless LAN Mesh Networks\n■\n399\nnode it is forming a peer relationship with to avoid joining rogue\nmesh nodes. This mutual authentication requirement is needed to\nprevent unauthorized mesh nodes from joining mesh networks.\n4.\nMesh node and mesh node authorization: A mesh node should\nauthorize the authenticating mesh node before forwarding traffic\nto and from it. The joining mesh node should also authorize its\npeer. By also obtaining authorization, both peers are assured that\nthe mesh node they are joining is authorized to perform the services\nof a mesh node.\n5.\nMesh node and mesh node key agreement: Two mutually authenti-\ncated mesh nodes should undergo handshakes to establish a fresh\nshared key to encrypt, authenticate, and integrity protect all traffic\nflowing between them. This key must be a short-lived key that is\nfreshly derived when the session is initiated and deleted once the\ncommunication between the mesh nodes is terminated.\n6.\nLocation privacy: Security between mesh node and client as well as\ntwo mesh nodes should be agnostic about location of the devices\nin question. Identities of mesh devices and clients should have no\ncorrelation with physical locations of those devices.\n7.\nSignaling authentication: Management and control frame protection\nis important in mesh backhaul as well as mesh access. Such broad-\ncast frames must be distinguishable from those announced by an\nattacker.\n8.\nService availability: A mesh node must be protected from DoS at-\ntacks and continue to offer services under such attacks. Even better\nis if such attackers can be located and mitigated in case of service\ndisruption. A mesh client cannot be excluded by a DoS attacker.\n9.\nSecure routing: Because multi-hop and multi-path routings are used\ninside, upstream, and downstream traffic forwarding from a wired\nportal, any routing protocol in operation must be secure against\nmalicious attacks.\n10.\nSecure MAC: The MAC protocol employed in mesh backhaul as well\nas access must be sufficiently resilient against RF and media-access\nattacks.\n11.\nSecure bridging: Because a mesh network can be interworked with\nother 802 LAN networks, any bridging protocol in use must be se-\ncure against any malicious attacks launched from those LAN net-\nworks.\nSome of the requirements above are discussed in detail in [17] and\nattempt to derive theoretical models of the attacks which may be launched\nwhen these requirements are not met by a WLAN mesh network.\nWe look at the proposals which were presented at TGs next.\n" }, { "page_number": 411, "text": "400\n■\nSecurity in Wireless Mesh Networks\n12.8\nSecurity in IEEE 802.11s WLAN Mesh\n12.8.1\nThe Original IEEE 802.11s Proposal\n12.8.1.1\nOverview\nThe original proposal uses the IEEE 802.11i concepts and mechanisms for\nmesh discovery and mesh association. It supports distributed and central-\nized models for AS functions. It utilizes optional additional security mech-\nanisms to support scalable security for data and management traffic.\nScalable security for data and management traffic allows pre-shared mul-\nticast keys so that information may be broadcast to all neighbors of an\nMP. IEEE 802.11i mechanisms are used to distribute the required 802.11i\nkeys and optional keys. These multicast keys are either unique to each\nMP (Neighbor Master Keys [NMK] and Neighbor Temporary Keys [NTK]) or\npre-shared among all MPs (Group Master Key [GMK] and Group Temporary\nKeys [GTK]).\nIEEE 802.11i required keys are pair-wise keys for securing the link be-\ntween a client and AP (PMK, PTK) and group keys for all nodes (GTK).\nThe optional keys for mesh networks are local multicast keys and global\nmulticast group keys (MMK/MTK). The local multicast keys support one\nkey per neighbor transmitting the data. The global multicast group keys\nsupport one multicast encryption key per multicast group.\nBasic 802.11i functions are extended to provide multi-hop encryptions\nfor unicast and multicast data or control frames. The extensions occur at\nthe neighbor security associations in mesh beacon or neighbor discovery\nHello functions.\n12.8.1.2\nSecurity Framework\nThe original IEEE 802.11 TGs security proposal is based on 802.11i RSNA\nsecurity and supports both centralized and distributed IEEE 802.1x-based\nauthentication and key management. In a WLAN mesh, an MP performs\nboth the supplicant and the authenticator roles, and may optionally perform\nthe role of an AS. The AS may be co-located with an MP or be located in\na remote entity to which the MP has a secure connection (this is assumed\nand not specified by the 802.11s proposal). Figure 12.2 shows the security\nframework in a WLAN mesh network. A node establishes RSNA in one of\nthree ways:\n1.\nCentralized 802.1x authentication model\n2.\nDistributed 802.1x authentication model\n3.\nPre-shared key authentication model\nThe first two use 802.1x EAP-based authentication followed by an\n802.11i-based four-way handshake. A central AS is used in the first model\nwhereas it is presumed that each MP in the MP–MP perform mutual\n" }, { "page_number": 412, "text": "Security in Wireless LAN Mesh Networks\n■\n401\nSecure connection\nEAP\nauthentication\nEAP\nauthentication\nEAP\nauthentication\nAS\nMP\nMAP\nSTA\nEAPOL-start\nEAPOL-start\nEAPOL-start\n4way handshake\n4way handshake\n4way handshake\nAuthenticator\nSupplicant\nFigure 12.2\nExample security exchanges in WLAN mesh.\nauthentication in the second model. The pre-shared model, where a sin-\ngle key is shared among the mesh does not quite scale to mesh networks\nwhere multi-hop routing is required. In particular, it is infeasible to secure\nrouting functionality when a pre-shared key is used in a mesh with more\nthan two nodes, because it is no longer possible to reliably determine the\nsource of any message. Alternatively, each MP may be provisioned with its\nunique pre-shared key, but then this also presents an unscalable model as\nevery MP must be provisioned with all of the MPs in the mesh.\nIEEE 802.11 TGs is effectively taking a different approach to solving the\nWLAN mesh security. At the time of writing this chapter, there were two\nproposals which were presented at the IEEE 802 Plenary meeting at San\nDiego, California, in July 2006.\n12.8.2\nCurrent IEEE 802.11s Security Proposals\nAt the time of publication, two security proposals were evaluated by TGs;\nsince, the core of Intel’s proposal has been adopted into the TGs base\nspecification though many security issues still remain to be stabilized. Both\nproposals are preceded by an almost common security framework. We first\ndiscuss that framework.\n" }, { "page_number": 413, "text": "402\n■\nSecurity in Wireless Mesh Networks\n■\nDiscovery: Each MP advertises its security policy in the beacons\nand probe responses it generates. Other MPs within range interpret\nreceived beacons and probe responses to learn the security policy\nof the message source.\n■\nIEEE 802.11 authentication: When used, this performs peer authen-\ntication and implicit authorization to perform mesh forwarding.\n■\nRole determination: The security policy is determined by an algo-\nrithm that also determines which party plays the role of IEEE 802.1X\nauthenticator and which plays the role of supplicant for each link\ninstance. This algorithm executes prior to beginning the link estab-\nlishment procedure.\n■\nLink security policy selection: This involves the supplicant selecting\namong the pairwise cipher suites and authenticated key manage-\nment protocols advertised by the authenticator in its beacons and\nprobe responses. The supplicant asserts its selection through the\nWLAN mesh link establishment procedure. The IEEE 802.1x entity\ncloses its controlled port when the secure link establishment proce-\ndure begins.\n■\nAuthentication and key management: After link establishment is as-\nserted, the authenticator initiates IEEE 802.1X authentication fol-\nlowed by a variant of the authenticated key management process\ndefined in Clause 8.5 of [6] to enable the authenticator and suppli-\ncant to mutually authenticate and establish fresh keys to secure the\n802.11 link. IEEE 802.1X authentication may be null if a pre-shared\nkey is optionally employed.\n■\nSecure link operation: Once authenticated key management com-\npletes successfully, the IEEE 802.1X entity opens its controlled port\nto allow data to flow, which is now protected.\nWhen security is enabled, mutual authentication between the two par-\nties must be achieved and thus at least one of IEEE 802.11 authentication\nor authentication and key management is required.\n12.8.2.1\nProposal from Intel Corporation\nOne of the two proposals originates from a group of security researchers\nfrom Intel Corporation. The proposal leverages IEEE 802.11i to secure the\nmesh transport and is summarized in this section.\nWhen a mesh node wants to utilize IEEE 802.1X to authenticate and\nauthorize with other MPs, it shall advertise its security policy by including\nthe RSN information element into its beacons and probe responses. An MP\nshall also set bits 7 and 8 of the RSN Capabilities field in the RSN information\nelement as follows:\n" }, { "page_number": 414, "text": "Security in Wireless LAN Mesh Networks\n■\n403\n■\nBit 7: The mesh node shall set this bit to 1 if it uses the mesh default\nrole determination scheme. Otherwise, the node shall set this bit\nto 0 if it uses some other role determination scheme, such as a\nproprietary scheme. The specification of other schemes is outside\nthe scope of this proposal and the TGs standard.\n■\nBit 8: This bit is meaningful when bit 7 is set to 1. The mesh node\nshall also set bit 8 to 1 if the mesh node can execute the role of\nthe IEEE 802.1X authenticator; otherwise, it sets this bit 0. Because\na mesh node must relay on an authentication database, it must ei-\nther provision it locally or be able to reach an 802.1X authentication\nserver. Thus, if either case is true, then bit 8 may be set to 1; other-\nwise setting this bit to 0 indicates that this mesh node has no access\nor means to 802.1X authenticate its peers.\nWhen an MP wishes to use 802.1X for authentication and authorization\nof different mesh roles, it inspects beacons and probe responses from the\nother MPs. When it receives a beacon or probe response from another MP,\nthe receiving MP shall examine whether bit 7 of the Capabilities field of the\nRSN information element from the message is set to 1. If both MPs have\nadvertised the ability to employ the proposed role determination by both\nsetting bit 7 to 1, then the proposed standard is employed. Otherwise, if\none of the mesh peers has not set bit 7 to 1, then based on the MPs policy,\na non-standard role determination may be negotiated or otherwise the MPs\nfail to establish a secure link.\nIf an IEEE 802.1X-based authentication and key management method is\nused, the MP playing the role of the IEEE 802.1X supplicant shall include\nan RSN information element in the association request specified by this\nmechanism. In the RSN information element, the supplicant MP shall specify\nexactly one pairwise cipher suite and one authenticated key management\nsuite.\nIn a wireless mesh network, all mesh nodes must utilize the same group\ncipher suite. Therefore, a supplicant MP must include the same group ci-\npher suite as advertised by the other MPs, especially the authenticator MP;\nsimilarly, the supplicant MP shall reject association requests from the au-\nthenticator MP (with status code 41), if the group cipher suite advertised\nby the authenticator MP does not match its own.\nThe authenticator MP shall also reject the association request from the\nsupplicant MP if either the pairwise cipher suite (with status code 42) or\nauthenticated key management suite (with status code 43) selected by\nthe supplicant is not included in the corresponding lists of pairwise ci-\npher suites and authenticated key management suites specified in its own\nbeacons and probe responses. The authenticator MP may also reject the\nsupplicant MP’s association request for other reasons unrelated to security.\nThe authenticator MPs may accept the association request if the supplicant\n" }, { "page_number": 415, "text": "404\n■\nSecurity in Wireless Mesh Networks\nselected pairwise and authenticated key management suites from among\nthose specified by the authenticator in its beacons and probe responses.\nOnce the role of the supplicant and authenticator is established between\ntwo MPs, the logic followed for the security negotiation, 802.1X authenti-\ncation, and key establishment is the same as that defined in IEEE 802.11i.\nThe proposal provided by Intel allows for as much of the re-use of IEEE\n802.11i with the modifications and enhancements to include the role and\nauthorizations of the peers to behave as mesh nodes.\n12.8.2.2\nProposal from Tropos Networks and Earthlink\nThe second proposal is called Comminus, jointly proposed by Tropos®\nNetworks and Earthlink. This proposal attempts to provide peer authenti-\ncation prior to full authorization and key management. Comminus attempts\nto partition the steps of authentication, authorization, and secure link estab-\nlishment as a means to allow flexibility in requiring access to an authenti-\ncation server or provisioning of a full authentication database. By using the\nstandard 802.11 authentication mechanism versus 802.1X, Comminus obvi-\nates the need to negotiate the supplicant and authenticator roles. Comminus\nbegins with the requirements of dynamically generating ephemeral session\nkeys, not being susceptible to active or passive attacks, ability to provide\nsome level of DoS resistance, providing implicit or no authorization, and\nproviding authorization as an overlay.\nComminus protocol is based on SKEME [18], a well-known key agree-\nment protocol that is known to be secure. It is based on Diffie–Hellman and\nto achieve mutual authentication can employ pre-shared keys or certificate-\nbased authentication. The Diffie–Hellman authenticated key agreement uses\nthe 802.11 authentication frames and can provide mutual authentication\nbetween two nodes (no notion of supplicant or authenticator or need of\nan AS). Comminus provides perfect forward secrecy. However, to achieve\nsuch mutual authentication, each MP must now be provisioned with all of\nits peer MPs’ pre-shared keys or a means to validate their certificates, if\nprovided. Without the use or means to authenticate such credentials, e.g.,\npre-shared keys or public keys (e.g., certificates), the result is only in a\nsecure key agreement with two unauthenticated parties. That is, there are\nassurances that there is no MIM but no gains on authentication. Lastly, there\nis no means to complete the authorization between the two MPs. However,\nonce a key has been secured among the two MPs, though maybe lacking\nin authentication and authorization, it can provide the following additional\nproperties:\n1.\nResistant to passive and DoS attacks, limited active attacks possible\n2.\nMay allow using ephemeral keys for management frame protection\nafter authentication is complete\n3.\nNo authorization, it is to be used for mesh formation only\n" }, { "page_number": 416, "text": "Security in Wireless LAN Mesh Networks\n■\n405\n4.\nNo master key exposure issues as shared secret is known by only\ntwo nodes\nThe Diffie–Hellman computation is generally expensive to perform in\nhardware even though there are optimized versions of the algorithm now\navailable. There is also no real re-authentication or key refresh mechanisms\nbuilt into Comminus, nor is there a means to address mobility. Further, it\nis not clear about the lifetime of the session keys in case of link or node\noutage between the two nodes sharing the same secret. Comminus does not\nprovide a full WLAN mesh security solution. Hence, it proposes to use EAP\nmethods along with AAA/RADIUS for a mesh node to servicing additional\nmesh functionalities, such as routing and bridging on top of mesh link\nformation.\n12.9\nDiscussion and Conclusion\nSecurity is often an afterthought in new technology evolutions. But to make\nthese technologies a commercial success, security problems need to be\nsolved up front with careful considerations into topics like authentication,\nauthorization, and access control of all members of the network; data and\nmanagement frame confidentiality, privacy, authenticity, and integrity; in-\ntrusion detection and prevention; rogue member detection and preven-\ntion; malicious attack detection and prevention; and damage containment\nand mitigation plans. Especially for multi-hop wireless networks, e.g., a\nWLAN mesh, it is necessary to address end user concerns over these require-\nments. This chapter discussed many security issues, threats, and solution\napproaches for WLAN mesh networks with some highlights of the current\nsecurity proposals discussed within the IEEE 802.11 TGs. Further, there are\nopen issues that remain:\n■\nCentralized AAA and AS schemes are not scalable in WLAN mesh\nnetworks.\n■\nThere is no single efficient and reliable security solution suitable for\nWLAN mesh as many of those solutions may be compromised due\nto vulnerabilities of channels and nodes in shared media, absence\nof reliable links to infrastructure, and dynamic topology changes.\n■\nAttackers may launch MIM and modification attacks against routing\nprotocols, such as AODV and OLSR.\n■\nWithout strong authorization, attackers may enter into the network\nand impersonate legitimate nodes and not follow protocol rules.\n■\nAttackers may create sinkholes, black holes, gray holes, and worm-\nholes to disrupt network traffic and take shortcuts.\n■\nGreedy nodes may utilize MAC back-off procedures and NAV for vir-\ntual carrier sense mechanisms of 802.11 MAC and cause congestions\nin the network.\n" }, { "page_number": 417, "text": "406\n■\nSecurity in Wireless Mesh Networks\n■\nAvailability of an AS and mechanisms to authenticate in its exchanges\nusing peer-based mutual authentication schemes need security anal-\nysis for WLAN mesh.\n■\nGroup key management remains a challenge in the absence of a\ncentral authority, trusted third party, or server to manage the keys.\nSome distributed and self-organizing key management schemes may\nbe needed for WLAN mesh.\nMost WLAN mesh security technologies (inclusive of the ones proposed\nat IEEE TGs) are attempting to leverage existing EAP and IEEE 802 se-\ncurity mechanisms and embed mesh-specific extensions as needed. How-\never, techniques for security monitoring, response systems to detect attacks,\nmonitoring service disruption, responding quickly to attacks, and mitigat-\ning/containing damage in WLAN mesh networks are still limited [24,25].\nTGs focuses only in addressing the link and network layer security prob-\nlems as it presumes use of other security mechanisms such as IPSec, VPN,\nand other technologies for securing the higher layers. Unfortunately, there\nis very little focus on cross- and multi-layer coordinated security protocols\nto combat simultaneous attacks on different protocol layers. Much work\nremains to develop a framework for building systems that can actually bat-\ntle multi-protocol attacks as well as detect and prevent intrusion in WLAN\nmesh networks.\nReferences\n[1]\nIEEE P802.11s/D0.01, Amendment X: ESS Mesh Networking, IEEE, Draft\nStandard, March 2006, work in progress.\n[2]\nhttp://www.ietf.org/html.charters/manet-charter.html\n[3]\nhttp://www.ietf.org/html.charters/nemo-charter.html\n[4]\nR. Ogier, F. Templin, and M. Lewis, Topology Dissemination Based on\nReverse-Path Forwarding, RFC 3684, IETF, February 2004.\n[5]\nI. Chakeres, E. Belding-Royer, and C. Perkins, Dynamic MANET On-\nDemand (DYMO) Routing, draft-ietf-manet-dymo-03, IETF, Internet Draft,\nOctober 2005, work in progress.\n[6]\nIEEE Std 802.1X-2004, 802.1X: Port-Based Network Access Control, IEEE,\nLAN/MAN Standard, 2004.\n[7]\nIEEE Std 802.11i-2004, 802.11i: Amendment 6: Medium Access Control\n(MAC) Security Enhancements, IEEE, LAN/MAN Standard, 2004.\n[8]\nD. Jones, Metro-Mesh: A Hacker’s Paradise, May 2006, available at\nhttp://www.darkreading.com/document.asp?doc id=95609\n[9]\nJ. Edney and W.A. Arbaugh, Real 802.11 security: Wi-Fi protected access\nand 802.11i, Addison-Wesley Reading, MA, 2004.\n[10]\nhttp://www.openssh.org\n[11]\nhttp://www.openssl.org\n" }, { "page_number": 418, "text": "Security in Wireless LAN Mesh Networks\n■\n407\n[12]\nC. Perkins, E. Belding-Royer, and S. Das, Ad hoc On-Demand Distance\nVector (AODV) Routing, RFC 3561, IETF, July 2003.\n[13]\nM. Zapata and N. Asokan, Securing Ad hoc Routing Protocols, ACM Work-\nshop on Wireless Secuirty (WiSe), September 2002.\n[14]\nK. Sanzgiri, B. Dahill, B.N. Levine, C. Shields, and E.M. Belding-Royer, A\nSecure Protocol for Ad hoc Networks, IEEE International Conference on\nNetwork Protocols (ICNP), 2002.\n[15]\nY. Hu, A. Perrig, and D. Johnson, Ariadne: A Secure On-demand Routing\nProtocol for Ad hoc Networks, ACM Annual International Conference on\nMobile Computing and Networking (MOBICOM), September 2002.\n[16]\nB. Aboba, L. Blunk, J. Vollbrecht, J. Carlson, and H. Levkowetz, Eds., Ex-\ntensible Authentication Protocol (EAP), RFC 3748, IETF, June 2004.\n[17]\nY. Zhang and Y. Fang, ARSA: An attack-resilient security architecture for\nmulti-hop wireless mesh networks, IEEE Journal on Selected Areas in Com-\nmunications, 4th Quarter, 2006.\n[18]\nH. Krawczyk, SKEME: A Versatile Secure Key Exchange Mechanism for the\nInternet, August 1995.\n[19]\nIEEE P802.11w/D0.0, Amendment 11: Protected Management Frames, IEEE,\nDraft Standard, March 2006, work in progress. TGs has since progressed\nand, as of this publication is working towards a new draft version 2.0.\n[20]\nIEEE Std 802.3-2002, 802.3: Carrier Sense Multiple Access with Collision\nDetection (CSMA/CD) Access Method and Physical Layer Specifications,\nIEEE, LAN/MAN Standard, 2002.\n[21]\nN. Borisov, I. Goldberg, and D. Wagner, Intercepting Mobile Communica-\ntions: The Insecurity of 802.11, ACM Annual International Conference on\nMobile Computing and Networking (MOBICOM), September 2002.\n[22]\nL. Buttyan and J.-P. Hubaux, Report on a working session on security in\nwireless ad hoc networks, ACM Mobile Computing and Communications\nReview, 7, 1, 2002.\n[23]\nV. Gupta, S. Krishnamurthy, and M. Faloutsos, Denial of Service Attacks at\nthe MAC Layer in Wireless Ad hoc Networks, IEEE Military Communication\nConference (MILCOM), 2002.\n[24]\nH. Yang, H. Luo, F. Ye, S. Lu, and L. Zhang, Security in Mobile Ad hoc\nNetworks: Challenges and Solutions, IEEE Wireless Communications, 11,\n1, 3847, 2004.\n[25]\nJ.-P. Hubaux, L. Butttan, and S. Capkun, The Quest for Security in Mo-\nbile Ad hoc Networks, ACM International Symposium on Mobile Ad Hoc\nNetworking and Computing (MOBIHOC), 2001.\n[26]\nS. Fluhrer, I. Mantin, and A. Shamir, Weaknesses in the Key Scheduling\nAlgorithm of RC4, Eighth Annual Workshop on Selected Areas in Cryptog-\nraphy, 2001.\n[27]\nP. Oechslin, Making a Faster Cryptanalytic Time-Memory Trade-Off, Pro-\nceedings of Crypto, 2003.\n[28]\nS. Cam-Winget, R. Housley, D. Wagner, and J. Walker, Security flaws in\n802.11 data link protocols, Communications of the ACM, Volume 46, Issue\n5, May 2003.\n[29]\nIEEE STD 802.11-2007. Wireless Local Area Networks, IEEE, WLAN Stan-\ndard, 2002.\n" }, { "page_number": 419, "text": "" }, { "page_number": 420, "text": "Chapter 13\nSecurity in IEEE 802.15.4\nCluster-Based Networks\nMoazzam Khan and Jelena Misic\nContents\n13.1\nCluster-Based Networks and Network Lifetime ................... 412\n13.2\nSecurity in Wireless Sensor Networks ............................. 414\n13.2.1\nSecurity Techniques ....................................... 414\n13.2.1.1 Data Confidentiality ............................. 414\n13.2.1.2 Data Authentication ............................. 414\n13.2.2\nData Integrity .............................................. 415\n13.2.3\nReplay Protection.......................................... 415\n13.3\nOverview of IEEE 802.15.4 Security Operations .................. 415\n13.3.1\nAddressing ................................................. 415\n13.3.1.1 Outgoing Frame Packet and Use of ACL ....... 417\n13.3.1.2 Incoming Frame Packet and Use of ACL ....... 417\n13.3.2\nNo Security ................................................ 418\n13.3.3\nAES-CTR ................................................... 418\n13.3.4\nAES-CBC-MAC ............................................. 419\n13.3.5\nAES-CCM................................................... 419\n13.3.6\nReplay Protection.......................................... 419\n13.4\nKey Management Models .......................................... 419\n13.4.1\nProbabilistic Keying Models .............................. 420\n13.4.2\nDeterministic Keying Models ............................. 421\n13.4.3\nHybrid Keying Models .................................... 421\n409\n" }, { "page_number": 421, "text": "410\n■\nSecurity in Wireless Mesh Networks\n13.4.4\nGroupwise Keying Models ................................ 421\n13.4.5\nKey Updates ............................................... 422\n13.4.5.1 Static Keying Schemes........................... 422\n13.4.5.2 Dynamic Keying Schemes....................... 423\n13.4.6\nLimitations of IEEE 802.15.4 Standard from\nthe Security Aspect ........................................ 423\n13.5\nSecurity Services Provided by ZigBee Alliance.................... 423\n13.5.1\nKeyed Hash Function for Message Authentication ....... 424\n13.5.2\nSymmetric-Key Key Establishment Protocol.............. 424\n13.5.2.1 Exchange of Ephemeral Data ................... 425\n13.5.3\nGeneration of Shared Secret .............................. 426\n13.5.4\nDerivation of Link Key .................................... 427\n13.5.5\nConfirming Link Key ...................................... 428\n13.5.6\nCommunication Steps in SKKE Protocol ................. 428\n13.6\nSummary............................................................ 431\nReferences................................................................. 432\nThe recently adopted IEEE 802.15.4 standard is poised to become the key\nenabler for low complexity, ultra-low power consumption, low data rate\nwireless connectivity among inexpensive devices such as sensors. This stan-\ndard will play an important role in sensitive applications including habitat\nmonitoring, burglar alarms, inventory control, medical monitoring, emer-\ngency response, and battlefield management which needs reliable and se-\ncure data transfer.\nTwo network topologies are allowed by the standard, but both of them\nrely on the presence of a central controller device known as the PAN coor-\ndinator. In the peer-to-peer topology, devices can communicate with one\nanother directly, as long as they are within the physical range. In star-based\ntopology, the devices must communicate through the PAN coordinator. The\nnetwork uses two types of channel access mechanism: one based on a\nslotted CSMA-CA algorithm in which the slots are aligned with the beacon\nframes sent periodically by the PAN coordinator, and another based on un-\nslotted CSMA-CA in which there are no beacon frames. The beacon-enabled\nmode and the star-based1 hierarchical topology appear to be better suited\nto sensor network implementation than their peer-to-peer counterparts be-\ncause the PAN coordinator can act as both the network controller and the\nsink to collect the data from the sensor nodes. Within one cluster, time is\norganized in superframes which are delineated by beacons sent by the PAN\n1 In the text that follows we will refer to star-based topology as cluster-based topology.\n" }, { "page_number": 422, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n411\ncoordinator. A superframe is further organized in active part, where nodes\ncan transmit using CSMA-CA or TDMA (called guaranteed time slots), and\ninactive part, where all nodes sleep. Larger areas under surveillance can be\nefficiently covered by interconnecting clusters in mesh topology through\ntheir coordinators. This feature is enabled through the existence of the in-\nactive superframe part because the coordinator can then switch to another\ncluster and communicate as an ordinary node. When communication in a\nforeign cluster is finished, the coordinator returns to its own cluster.\nWireless devices used for sensing the environment are low in com-\nputational power and memory resources. The bandwidth offered by IEEE\n802.15.4 standard is low, because the standard allows the PAN to use ei-\nther one of three frequency bands: 868 to 868.6, 902 to 928, and 2400 to\n2483.5 MHz with raw data rates of 20, 40, and 250 kbps, respectively. How-\never the bandwidth available to the application is further decreased due to\nCSMA-CA access with small back-off windows (default back-off window\nsizes without power saving mode are 8, 16, 32, 32, 32, respectively, for five\nallowed back-off attempts). Also, in downlink communications, the PAN\ncoordinator first has to advertise the packet in the beacon, then the node\nhas to send the request packet asking for downlink transmission, and fi-\nnally, downlink transmission can commence. Therefore, in the presence of\nmany nodes in the cluster, effective bandwidth left to the application is less\nthan 20 percent of the raw bandwidth [15].\nProviding security services in such wireless sensor networks is a techni-\ncal challenge. Algorithms for key exchange which naturally include authen-\ntication elements and addition of packet signature will further decrease the\nbandwidth available to the sensing application. Besides, complex compu-\ntations often involved in public key cryptography might consume too much\nenergy and memory resources. Therefore, the goal of designing low-power\nsensor devices forces security mechanism to fit under processing, memory,\nand bandwidth constraints.\nThis chapter is organized as follows. In Section 13.1, we explain the\nrelationship between the sensor network architecture and its availability\nfor both data collection and event sensing applications. We believe that\nnetwork availability for sensing applications has the same importance as\ndata integrity and to some extent data confidentiality. Section 13.2 ex-\nplains the need of security in wireless sensor networks and which types\nof security techniques are considered in such networks. A detailed de-\nscription of security features of IEEE 802.15.4-based [3] sensor networks is\npresented in Section 13.3. Section 13.4 discusses keying models currently\nused in WPANs. Security issues addressed by the ZigBee alliance speci-\nfications [4] are discussed in Section 13.5. Finally, Section 13.6 concludes\nthis chapter.\n" }, { "page_number": 423, "text": "412\n■\nSecurity in Wireless Mesh Networks\n13.1\nCluster-Based Networks and Network Lifetime\nOne of the most significant benefits of sensor networks is that they ex-\ntend the computation capability to physical environments where human\nbeings cannot reach. However, energy possessed by sensor nodes is lim-\nited, which becomes the most challenging issue in designing sensor net-\nworks. The main power consumptions in sensor networks are computation\nand communication between sensor nodes. In particular, the ratio of energy\nconsumption for communication and computation is typically in the scale\nof 1000 [12]. Therefore it is critical to enable collaborative information pro-\ncessing and data aggregation to prolong the lifetime of sensor networks.\nThe choice of network topology in wireless sensor networks is still an\nopen question. However, it seems that the choice of topology is an issue of\ntrade-off between node simplicity and homogeneity versus the duration of\nnetwork lifetime. For sensor networks covering large geographic areas, it\nis difficult to replace sensor batteries when they are exhausted, and there-\nfore when nodes close to the sink die the whole network is unavailable.\nTherefore, from the aspect of availability, long network lifetimes become\nan important security aspect.\nWireless sensor networks can carry two different types of sensing. The\nfirst kind of sensing is data collection where nodes in the network fre-\nquently communicate to report measurements that lead to continuous flow\nof data from nodes. Depending on the application requirements, some\ncollective sleep technique for all the nodes in the cluster can be used to\nextend the network lifetime. Data collection applications exploit spatial\ncorrelation of sensed data and, to save bandwidth, perform some kind of\ndata aggregation. In peer-to-peer IEEE 802.15.4 architectures, aggregation\nis performed in nodes which are conveying sensed data toward the sink.\nIn cluster-based architectures, aggregation occurs at the PAN coordinator\nand aggregated packets are conveyed to the next coordinator along the\npath, possibly over a more powerful link (GTS) compared to the link type\nwhich is available to ordinary nodes (CSMA-CA). From the aspect of avail-\nable bandwidth, the presence of GTS links between the cluster coordinators\ngives the cluster-based networks an advantage over the peer-to-peer net-\nworks. Also, the aggregation done by the coordinator can be made much\nmore secure than the aggregation in peer-to-peer networks because the\ncoordinator is always aware of the identities of the nodes which partici-\npate in the aggregation (because this is done in the attachment process),\nwhile the set of neighbors in the peer-to-peer network might depend on\nthe type of query. From the aspect of lifetime, it is reasonable to assume\nthat PAN coordinators will have higher power resources than the ordinary\nnodes, which, combined with the GTS access, will extend the lifetime of\nthe network (because they will relay packets).\n" }, { "page_number": 424, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n413\nIn the second kind of sensing, communication occurs only when some\nimportant event occurs and data is communicated in bursty fashion from\nnodes toward the sink. For applications where event detection is the target\n(e.g., enemy troops movement, detection of noise level), sensors are re-\nquired to be vigilant most of the time, which means that collective sleep of\nthe nodes is prohibited. Event detection requires reporting only when an\nevent occurs in contrast to data collection where communication of mea-\nsurements is more frequent. In this case aggregation is avoided and it is\nimportant to deliver the sensed data to the sink within some time bound\n(time bounds are not important for data collection due to time correlation\nof sensed data). In event-detection applications, network availability and\ndata integrity are much more critical than in data collection applications.\nAgain, we argue that a cluster-based architecture where PAN coordinators\nhave higher power resources, GTS links for communication, and reliable in-\nformation about cluster members offers better availability and data integrity\nthan a peer-to-peer architecture.\nNodes in wireless sensor networks can directly communicate with nearby\nnodes. Nodes that are not within direct communication range use other\nnodes to relay messages between them. Routing in such a multi-hop net-\nwork is challenging due to the lack of central control and the high dynamics\nof the network. Recent work has focused on discovering and maintaining\nroutes that keep the connectivity between the nodes or that minimize the\nnumber of hops on a path. One important restriction of a wireless sensor\nnetwork is that nodes are energy-constrained as they are normally powered\nby batteries. However, the algorithms that aim to minimize the path length\nmay ignore fairness in routing, for example, the shortest-path routing is\nlikely to use the same set of hops to relay packets for the same source and\ndestination pair. This will heavily load those nodes on the path even when\nother feasible paths exist. Such an uneven use of the nodes may cause\nsome nodes to die earlier, thus creating holes in the network, or worse,\nleaving the network disconnected.\nLow available bandwidth to nodes, CSMA-CA access, data aggregation,\nand routing in wireless sensor networks based on IEEE 802.15.4 make the\nimplementation of security a technical challenge. Even at the MAC layer it is\npossible to launch a denial-of-service attack which will drastically increase\nthe number of collisions and prevent data communication (due to CSMA-\nCA access and small back-off windows). The processing, communication,\nand aggregation cost of secure packets first increases both computational\nand communication overhead. To decrease this overhead all the security\nparameters and keying models under which the network will work are\nselected with great care so that the objectives of both secure communication\nand longer network life are achieved. These two objectives are competing\nand trade-off between them is necessary. For implementation of secure\n" }, { "page_number": 425, "text": "414\n■\nSecurity in Wireless Mesh Networks\nsensor network we have to compromise on network life to some extent\nand vice versa.\n13.2\nSecurity in Wireless Sensor Networks\nRadio is a shared medium; everything that is transmitted or received over a\nwireless network can be intercepted in such an environment. An adversary\ncan gain access to information by monitoring the communication among\nnodes. For example, few wireless receivers placed outside a house might\nbe able to monitor the light and temperature sensor readings of a sensor\nnetwork inside the house, thus revealing detailed information about the\noccupant’s daily personal activity. Similarly, an attacker can obtain a com-\nmodity sensor node and present it as a legitimate node inside the network;\nonce an attacker has a few nodes like that in a network, he can launch\na different types of attack, for example, denial of service, falsification of\nsensed data, dropping of sensed data, etc.\n13.2.1\nSecurity Techniques\nDifferent security techniques are employed to safeguard threats of such\neavesdropping, and we will discuss such techniques next.\n13.2.1.1\nData Confidentiality\nAll nodes in a sensor network communicate through one wireless medium,\nand listening to this medium is easy. Hence a network should not leak sen-\nsor data to any neighboring network or any node that is not part of the net-\nwork. The standard approach for keeping sensitive data secret is to encrypt\nthe data with a secret key that is carried by the intended receivers only.\n13.2.1.2\nData Authentication\nData authentication allows the receiver to verify that the data was really\nsent by the claimed sender. Authentication also prevents an attacker from\nmodifying a hacked device to impersonate another device. Because an ad-\nversary can easily inject messages, the receiver needs to ensure that the\ndata used in any decision-making process originates form a trusted source.\nData authentication is usually achieved through a symmetric mechanism\nwhere sender and receiver share a key to compute the Message Authenti-\ncation Code (MAC). The data is appended along with its MAC, and once\nthe receiver gets the data, it recalculates the MAC. If the same MAC is cal-\nculated that it received from same sender, it shares the key. Authentication\ncan be achieved both at the cluster level and the device level. Cluster-\nlevel authentication is achieved using a common network key, whereas\n" }, { "page_number": 426, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n415\ndevice-level authentication is achieved by using unique pairwise keys for\neach link in the network.\n13.2.2\nData Integrity\nData integrity allows the receiver to verify that the data received is the\nsame as the data sent by the sender and is not changed during its trans-\nmission to the receiver. If the MAC calculated by the receiver is the same\nas received, it means that the data was not altered during transmission to\nreceiver. Message authentication codes must be hard to forge without the\nsecret key. Consequently, if an adversary alters a valid message or injects\na bogus message, he will not be able to compute the corresponding MAC,\nand authorized receivers will reject these forged messages. In sensor net-\nworks data integrity is usually achieved in symmetric fashion and is again\nrelied on the appended MAC, hence integrity and authentication options\nallow trade-off between message protection and message overhead.\n13.2.3\nReplay Protection\nAn adversary that eavesdrops on a legitimate message sent between two\nauthorized nodes and replays it at some later time engages in replay attack.\nBecause the message originated from an authorized sender, it will have a\nvalid MAC, so the receiver will accept it again. Replay protection prevents\nthese types of attacks. The sender typically assigns a monotonically increas-\ning sequence number to each packet and the receiver rejects packets with\na smaller sequence number than it has already seen.\nIn symmetric mechanisms sender and receiver share one common key\nand rely on different security techniques for the secrecy of these keys.\nHence the whole security model revolves around the secrecy of symmetric\nkeys that can be either at the network level or a link level.\n13.3\nOverview of IEEE 802.15.4 Security Operations\nIEEE 802.15.4, a link layer security protocol, provides four basic security ser-\nvices: access control, message integrity, message confidentiality, and replay\nprotection. The security requirements can be tuned by setting the appro-\npriate control parameters of the protocol stack. If an application does not\nset any parameters, then security is not enabled by default. An application\nmust explicitly enable security features.\n13.3.1\nAddressing\nFor unique identification in a network or cluster, addressing in IEEE 802.15.4\nis accomplished via a 64-bit node identifier and a 16-bit network identifier.\n" }, { "page_number": 427, "text": "416\n■\nSecurity in Wireless Mesh Networks\nIEEE 802.15.4 supports a few different addressing modes. For example,\na 16-bit truncated address may be used in place of the full 64-bit node\nidentifier in certain cases. This allows the size of the source and destination\naddresses to vary between 0 and 10 bytes, depending on whether truncated\nor full addresses are used, and whether or not the node sends to broadcast\naddress.\nThe specification defines four packet types for the media access control\nlayer:\n1.\nBeacon packets\n2.\nData packets\n3.\nAcknowledgment packets\n4.\nControl packets\nThe specification does not support security for acknowledgment packets\nalthough security is optional for other packet types, depending on the\nneed of application. Depending on the threat environment, the application\nhas a choice of security suites that control the type of security protection\nprovided for the transmitted data. Each security suite offers a different set of\nsecurity properties and results in different packet formats. The IEEE 802.15.4\nspecification defines eight different security suites outlined in Table 13.1.\nWe can classify the suites by the properties they offer:\n■\nNo security\n■\nEncryption only (AES-CTR)\n■\nAuthentication only (AES-CBC-MAC)\n■\nEncryption and authentication (AES-CCM)\nTable 13.1\nSecurity Suites Supported by 802.15.4\nSecurity\nAccess\nData\nFrame\nIdentifier\nSuite Name\nControl\nEncryption\nIntegrity\nDescription\n0 × 00\nNone\n—\n—\n—\nNo security\n0 × 01\nAES-CTR\nX\nX\n—\nEncryption only\n0 × 02\nAES-CCM-128\nX\nX\nX\nEncryption and\n128-bit MAC\n0 × 03\nAES-CCM-64\nX\nX\nX\nEncryption and\n64-bit MAC\n0 × 04\nAES-CCM-32\nX\nX\nX\nEncryption and\n32-bit MAC\n0 × 05\nAES-CBC-MAC-128\nX\n—\nX\n128-bit MAC\n0 × 06\nAES-CBC-MAC-64\nX\n—\nX\n64-bit MAC\n0 × 07\nAES-CBC-MAC-32\nX\n—\nX\n32-bit MAC\n" }, { "page_number": 428, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n417\nAddress\nReplay\ncounter\nKey\nSecurity\nsuite \nFigure 13.1\nAccess control list entry. (From M. Khan, F. Amini, and J. Miˇsi´c, in\nMobile Ad-hoc and Sensor Networks, Springer, 2006. With permission.)\nThe specification supports MAC of sizes that can be either of 4, 8, or 16\nbytes long. The security feature of authentication is directly proportional to\nthe length of MAC and it is very difficult for an adversary to break or guess\na MAC of longer size. For example, with a 16-byte MAC, an adversary has\na 2−128 chance of forging the MAC. The trade-off is a larger packet size for\nincreased protection against authenticity attacks. The choice of secure au-\nthentication is tied with the addressing of devices in IEEE 802.15.4 devices.\nHence security suites are based on source and destination authentication\naddresses. Every device supporting IEEE 802.15.4 has an access control list\n(ACL) that controls what security suite and keying information is used by\neach device. Each device can support up to 255 ACL entries. Each entry\ncontains an 802.15.4 device address, a security suite identifier, and security\nmaterial as shown in (Figure 13.1).\nThe security material is the persistent state necessary to execute the\nsecurity suite. It consists of\n■\nCryptographic key\n■\nSecurity suite identifier\n■\nNonce state must be preserved across different packet encryption\ninvocations\n13.3.1.1\nOutgoing Frame Packet and Use of ACL\nIf security is enabled, the media access control layer looks up the destina-\ntion address in its ACL table. If there is a match ACL entry, the security suite\nand nonce specified in that ACL entry are used to encrypt or authenticate\nthe outgoing packets. On the other hand, in case of broadcast type of data\npacket where no specific destination address is mentioned, a default ACL\nentry is used, and this default entry matches all destination addresses.\n13.3.1.2\nIncoming Frame Packet and Use of ACL\nOn packet reception the media access control defined by IEEE 802.15.4 ex-\namines flag fields in the packet to determine if any security suite has been\napplied to that packet. If no security was applied, the packet is passed to\nan upper layer. Otherwise, the media access control layer finds an appro-\npriate ACL entry corresponding to the sender’s address. It then applies the\n" }, { "page_number": 429, "text": "418\n■\nSecurity in Wireless Mesh Networks\nFrame counter\nEncrypted MAC\nEncrypted payload\nKey counter\nFigure 13.2\nFrame format after adding security features. (From M. Khan, F.\nAmini, and J. Miˇsi´c, in Mobile Ad-hoc and Sensor Networks, Springer, 2006. With\npermission.)\nappropriate security suite and replay counter to the incoming packet. The\ngeneral structure of secured frame is shown in (Figure 13.2).\nWe will now provide more detail about the categories of security suites.\n13.3.2\nNo Security\nThis is the simplest security suite. Its inclusion is mandatory in all radio\nchips. It does not have any security material and operates as the identity\nfunction. It does not provide any security guarantees.\n13.3.3\nAES-CTR\nThis suite provides confidentiality protection using the AES (Advanced En-\ncryption Standard) block cipher with counter mode. To encrypt data under\ncounter mode, AES block cipher breaks the plaintext packet into 16-byte\nblocks p1....., pn and computes ci = pi ⊕E k(xi). Each 16-byte block uses\nits own varying counter, which we call xi. The recipient recovers the orig-\ninal plaintext by computing pi = ci ⊕E k(xi). Clearly the recipient needs\nthe counter value xi to reconstruct pi.\nThe xi counter, known as nonce or IV, is composed of\n■\na static flag field,\n■\nthe sender’s address, and\n■\nthree separate counters: a four-byte frame counter that identifies\nthe packet, a one-byte key counter field (the key counter is under\napplication control and can be incremented if the frame counter\never reaches its maximum value), and a two-byte block counter that\nnumbers the 16-byte blocks within the packet.\nThe requirement for employing infallible security is that the nonce\nmust never repeat within the lifetime of any single key, hence frame and\nkey counters are introduced to prevent nonce re-use. The two-byte block\ncounter ensures that each block will use a different nonce value.\nIn summary, the sender includes the frame counter, key counter, and\nencrypted payload into the data payload field of the packet as shown in\n(Figure 13.2).\n" }, { "page_number": 430, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n419\n13.3.4\nAES-CBC-MAC\nThis suite provides integrity protection using CBC-MAC. The sender can\ncompute either a 4-, 8-, or 16-byte MAC using the CBC-MAC algorithm,\nleading to three different AES-CBC-MAC variants. The MAC can only be\ncomputed by parties with the symmetric key. The MAC protects packet\nheaders as well as the data payload. The sender appends the plaintext data\nwith the MAC. The recipient verifies the MAC by computing the MAC and\ncomparing it with the value included in the packet.\n13.3.5\nAES-CCM\nThis security suite uses CCM mode for encryption and authentication.\nBroadly, it first applies integrity protection over the header and data pay-\nload using CBC-MAC, and then encrypts the data payload and MAC using\nAES-CTR mode. As such, AES-CCM includes the fields from both the authen-\ntication and encryption operations: a MAC and the frame and key counters.\nThese fields serve the same function as above. Just as AES-CBC-MAC has\nthree variants depending on the MAC size, AES-CCM also has three variants.\n13.3.6\nReplay Protection\nA receiver can optionally enable replay protection when using a security\nsuite that provides confidentiality protection. This includes AES-CTR and all\nof the AES-CCM variants. The recipients use the frame and key counter as a\nfive-byte value, the replay counter, with the key counter occupying the most\nsignificant byte of this value. The recipient compares the replay counter\nfrom the incoming packet to the highest seen, as stored in the ACL entry.\nIf the incoming packet has a larger replay counter than the stored one, then\nthe packet is accepted and the new replay counter is saved. If, however,\nthe incoming packet has a smaller value, the packet is rejected and ap-\nplication is notified of the rejection. We refer to this counter as the replay\ncounter, even though it is the same counter as the nonce, which is used for\nconfidentiality. The replay counter is not exposed to the application to use.\n13.4\nKey Management Models\nKey management is the process by which keys are generated, stored, pro-\ntected, transferred, updated, and destroyed. Keying refers to the process of\nderiving common secret keys among communicating parties. Pre-deployed\nkeying refers to the distribution of key(s) to the nodes before their deploy-\nment. Pairwise keying involves two parties agreeing on and communicating\nwith a session key after deployment, and group keying involves more than\ntwo parties using a common group key. Group keying is important for\nmulticasting.\n" }, { "page_number": 431, "text": "420\n■\nSecurity in Wireless Mesh Networks\nThe keying model that is most appropriate for an application depends\non the threat model that an application faces and what type of resources\nit is willing to expend for key management. Depending on application\ntypes, key management models can be discussed under the following pa-\nrameters: (1) network architectures such as distributed or hierarchical, (2)\ncommunication styles such as pairwise (unicast), groupwise (multicast),\nor networkwise (broadcast), (3) security requirements such as authenti-\ncation, confidentiality, or integrity, and (4) keying requirements such as\npre-distributed or dynamically generated pairwise, groupwise, or network-\nwise keys. The constrained energy budgets and the limited computational\nand communication capacities of sensor nodes make use of public cryp-\ntography impractical in large-scale sensor networks. At present, the most\npractical approach for bootstrapping secret keys in sensor networks is to\nuse pre-deployed keying in which keys are loaded into sensor nodes be-\nfore they are deployed. Several solutions based on pre-deployed keying\nhave been proposed in the literature, including approaches based on the\nuse of a global key shared by all nodes, approaches in which every node\nshares a unique key with the base station, and approaches based on ran-\ndom key sharing. In wireless sensor networks, nodes use pre-distributed\nkeys directly, or use keying materials to dynamically generate pairwise and\ngroupwise keys. The challenge is to find an efficient way of distributing\nkeys and keying materials to sensor nodes prior to deployment. Solutions\nto key distribution problems in WSN can use one of the following popular\napproaches.\n13.4.1\nProbabilistic Keying Models\nIn probabilistic solutions, keychains are randomly selected from a keypool\nand distributed to sensor nodes. For example, random pairwise key scheme\n[8] addresses unnecessary storage problems. In this scheme, each sensor\nnode stores a random set of N p pairwise keys to achieve probability p that\ntwo nodes are connected. At key setup phase, each node identity is matched\nwith N p other randomly selected nodes with probability p. A pairwise key is\ngenerated for each node pair, and is stored in every node’s keychain along\nwith the identity of its corresponding node. Similarly, [10] also proposed\nprobabilistic key pre-distribution scheme that relies on probabilistic key\nsharing among the nodes of a random graph and uses a simple shared-key\ndiscovery protocol for key distribution, revocation, and node re-keying.\nThis scheme showed that a pair of nodes may not share a key, but if\na path of nodes sharing keys pairwise exists between the two nodes at\nnetwork initialization, the pair of nodes can use that path to exchange a\nkey that establishes a direct link. Therefore, full shared-key connectivity\noffered by pairwise private key sharing between every two nodes becomes\nunnecessary.\n" }, { "page_number": 432, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n421\n13.4.2\nDeterministic Keying Models\nIn deterministic solutions, deterministic processes are used to design the\nkeypool and the keychains to provide better key connectivity. For example,\n[5] suggested that all possible link keys in a network of size N can be\nrepresented as an N × N key matrix. It is possible to store a small amount\nof information to each sensor node, so that every pair of nodes can calculate\na corresponding field of the matrix, and use it as the link key. Multiple space\nkey pre-distribution scheme [9] improves the resilience of Blom’s scheme. It\nuses a public matrix G and a set of ω private matrices D. Polynomial-based\nkey pre-distribution scheme [6] distributes a polynomial share (a partially\nevaluated polynomial) to each sensor node by using whichever pair of\nnodes can generate a link key.\n13.4.3\nHybrid Keying Models\nFinally, hybrid solutions use probabilistic approaches on deterministic so-\nlutions to improve scalability and resilience. Polynomial pool-based key\npre-distribution scheme [13] considers the fact that not all pairs of sen-\nsor nodes have to establish a key. It combines polynomial-based key pre-\ndistribution scheme [6] with the keypool idea in [8,10] to improve resilience\nand scalability.\n13.4.4\nGroupwise Keying Models\nIn hierarchical WSNs, sensor nodes require groupwise keys to secure mul-\nticast messages. One approach is to use secure but costly asymmetric cryp-\ntography [7], and IKA2 [17] use a Diffie–Hellman-based group key transport\nprotocol. Recently, some works on the public key cryptography protocols\n(e.g., elliptic curve cryptography) evaluation and efficiency measurements\non sensor node platforms showed optimistic results [11,18]. In a hierarchi-\ncal network, where a base station shares pairwise keys with all the sensor\nnodes, the base station can intermediate establishment of groupwise keys.\nLocalized Encryption and Authentication Protocol (LEAP) [19] provides a\nmechanism to generate groupwise keys which follow the LEAP pairwise\nkey establishment phase.\nAn important design consideration for security protocols based on sym-\nmetric keys is the degree of key sharing between the nodes in the system.\nAt one extreme, we can have networkwide keys that are used for encrypt-\ning data and for authentication. This key sharing approach has the lowest\nstorage costs and is very energy-efficient because no communication is re-\nquired between nodes for establishing additional keys. However, it has the\nobvious security disadvantage that the compromise of a single node will\nreveal the global keys. At the other extreme, we can have a key sharing\n" }, { "page_number": 433, "text": "422\n■\nSecurity in Wireless Mesh Networks\napproach in which all secure communication is based on keys that are\nshared pairwise between two nodes. From the security point of view, this\napproach is ideal because the compromise of a node does not reveal any\nkeys that are used by the other nodes in the network. However, under this\napproach, each node will need a unique key for every other node that it\ncommunicates with. Moreover, in many sensor networks, the immediate\nneighbors of a sensor node cannot be predicted in advance; consequently,\nthese pairwise shared keys will need to be established after the network\nis deployed. A unique issue that arises in sensor networks that needs to\nbe considered while selecting a key sharing approach is its impact on the\neffectiveness of in-network processing. Particular keying mechanisms may\nreduce the effectiveness of in-network processing.\nIEEE 802.15.4-compliant devices can share a network key such that each\ncluster shares only one key among all devices to exchange data and for\nauthentication purposes. This will ease the key management and memory\noverhead issues, but this comes at the cost of lower security. Similarly, IEEE\n802.15.4-compliant devices can also support pairwise key exchange that im-\nproves the overall security of a network where any two devices exchanging\ndata will share a different key. This improved robustness of network se-\ncurity comes at a cost, particularly in the overhead of key management. A\ndevice communicating with many devices in a network has to have differ-\nent keys for each corresponding communicating device, which will increase\nthe memory overhead on resource-scarce devices used in the network.\n13.4.5\nKey Updates\nKey management schemes are at the heart of securing such networks. Key\nmanagement schemes for sensor networks can be classified broadly into\nstatic and dynamic keying based on administrative key updates after net-\nwork deployment. While static schemes assume no updates, dynamic ones\nprovide for post-deployment key updates. The general security and perfor-\nmance objective of key management schemes include minimizing number\nof keys stored per sensor node, providing rich logical pairwise connectivity,\nand enhancing network resilience to node capture.\n13.4.5.1\nStatic Keying Schemes\nStatic keying management schemes (a.k.a. key-predistribution) perform key\nmanagement functions statically prior to or shortly after the deployment of\nthe network. Administrative keys are generated at the sensor manufacturing\ntime or by the base station upon network bootstrapping. Key assignment\nto nodes may be performed on a random basis or may take place based\non some deployment information. Once generated and assigned, keys are\npre-distributed to nodes. The main feature of static key management is the\n" }, { "page_number": 434, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n423\nfact that the above key management cycle takes place only once at or prior\nto initialization. Accordingly, lost keys due to node capture or failure are\nnot compensated.\n13.4.5.2\nDynamic Keying Schemes\nThe main feature of dynamic key management schemes is repeating the key\nmanagement process either periodically or on-demand to respond to node\ncapture. After initial keying, key generation, assignment, and distribution\nmight take place (in a process known as re-keying) to create new keys that\nreplace the keys assumed lost or revealed to an attacker so that the network\nis refreshed and the attacker loses information earned by node capture. An-\nother advantage of dynamic keying is that upon adding new nodes, unlike\nstatic keying, the probability of network capture does not necessarily in-\ncrease. Various dynamic key management techniques have been proposed\nwith different key management responsibility taken by different network\ncomponents.\n13.4.6\nLimitations of IEEE 802.15.4 Standard from\nthe Security Aspect\nHigher layers will determine when the security is to be used at the MAC\nlayer by any device and provide all keying material necessary to provide the\nsecurity services. Key management, device authentication, and freshness\nprotection may be provided by the higher layers, but is not addressed\nin IEEE 802.15.4 standard. The management and establishment of keys is\nthe responsibility of the implementer of higher layers. There is no simple\nway to group keys in IEEE 802.15.4-enabled WSNs because, as mentioned\nearlier, the ACL entries are only associated to a single destination address. A\ndetailed analysis of shortcomings of security features is mentioned by [16].\n13.5\nSecurity Services Provided by ZigBee Alliance\nAs explained above, the IEEE 802.15.4 addresses good security mecha-\nnisms, but it still does not address what type of keying mechanism will be\nused to employ supported security techniques.\nZigBee Alliance [4] is an association of companies working together\nto enable wireless networked monitoring and control products based on\nIEEE 802.15.4 standard. After the acceptance of 802.15.4 as IEEE standard,\nZigBee Alliance is mainly focused on developing network and application\nlayer issues. ZigBee Alliance is also working on application programming\ninterfaces (API) at the network and link layers of IEEE 802.15.4. The Al-\nliance also introduced secure data transmission in wireless sensor networks\n" }, { "page_number": 435, "text": "424\n■\nSecurity in Wireless Mesh Networks\nbased on IEEE 802.15.4 specification, but most of this work is in general\ntheoretical descriptions of security protocol at the network layer. There is\nno specific study or results published or mentioned by ZigBee Alliance\nin regard to which security suites perform better in different application\noverheads. ZigBee Alliance has recommended both symmetric and asym-\nmetric key exchange protocols for different networking layers. Asymmetric\nkey exchange protocols that mainly rely on public key cryptography are\ncomputationally intensive and their feasibility in wireless sensor networks\nis only possible with devices that are resource-rich both in computation\nand power.\n13.5.1\nKeyed Hash Function for Message Authentication\nA hash function is a way of creating a small digital fingerprint of any data.\nCryptographic hash function is a one-way operation and there is no prac-\ntical way to calculate a particular data input that will result in a desired\nhash value, thus it is difficult to forge. A practical motivation for construct-\ning hash functions from block ciphers is that if an efficient implementation\nof block cipher is already available within a system (either in hardware\nor in software), then using it as the central component for a hash func-\ntion may provide latter functionality at little additional cost. IEEE 802.15.4\nprotocol supports a well-known block cipher AES, and hence ZigBee Al-\nliance specification also relied on AES. ZigBee Alliance suggested the use\nof Matyas–Meyer–Oseas [14] as the cryptographic hash function that will be\nbased on AES with a block size of 128 bits.\nMechanisms that provide integrity checks based on a secret key are usu-\nally called MACs. Typically, message authentication codes are used between\ntwo parties that share a secret key to authenticate information transmitted\nbetween these parties. ZigBee Alliance specification suggests the keyed\nhash message authentication code (HMAC) as specified in the FIPS Pub\n198 [2]. A MAC takes a message and a secret key and generates a M ACtag,\nsuch that it is difficult for an attacker to generate a valid (message, tag)\npair and is used to prevent attackers forging messages. The calculation of\nMacT ag (i.e., HMAC) of data MacData under key MacK ey will be shown\nas follows:\nMacTag = MACMacKeyMacData\n13.5.2\nSymmetric-Key Key Establishment Protocol\nKey establishment involves two entities, an initiator device and a responder\ndevice, and is prefaced by a trust-provisioning step. Trust information (e.g.,\na master key) provides a starting point for establishing a link key and can\n" }, { "page_number": 436, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n425\nGenerate a challenge \nQEV and send to \ndevice U. \nCheck if a valid \nchallenge QEV within \ndomain D is received. \nIf a valid challenge\nQEU within domain D\nis received.\nDevice \nPAN \ncoordinator \nU⎥⎥ V⎥⎥ QEU⎥⎥ QEV\nU⎥⎥ QEU\nInitiator \nU \nResponder \nV \nFigure 13.3\nExchange of ephemeral data. (From M. Khan, F. Amini, and J. Miˇsi´c, in\nMobile Ad-hoc and Sensor Networks, Springer, 2006. With permission.)\nbe provisioned in-band or out-band. In the following explanation of the\nprotocol, we assume unique identifiers for initiator devices as U and for\nresponder device (PAN coordinator) as V. The master key shared among\nboth devices is represented as Mkey.\nWe will divide Symmetric-Key Key Establishment (SKKE) protocol\nbetween initiator and responder in the following major steps.\n13.5.2.1\nExchange of Ephemeral Data\nFigure 13.3 illustrates the exchange of the ephemeral data where the initia-\ntor device U will generate the challenge QEU. QEU is a statistically unique\nand unpredictable bit string of length challengelen by either using a ran-\ndom or pseudo-random string for a challenge Domain D. The challenge\ndomain D defines the minimum and maximum length of the challenge.\nD = (minchallengeLen, maxchallengeLen)\nInitiator device U will send the challenge QEU to a responder device\nwhich upon receipt will validate the challenge QEU by computing the bit-\nlength of bit string challenge QEU as Challengelen and verify that\nChallengelen ∈[minchallengelen, maxchallengelen]\nOnce the validation is successful, the responder device will also generate\na challenge QEV and send it to initiator device U . The initiator will also\nvalidate the challenge QEV as described above.\n" }, { "page_number": 437, "text": "426\n■\nSecurity in Wireless Mesh Networks\nZ = MACTag\nCalculate the shared secret Z\nMACTag = MACMkeyMACData\nResponder\nV\nInitiator\nU\nDevice\nPAN\ncoordinator\nU⎥⎥ V⎥⎥ QEU⎥⎥ QEV\nU⎥⎥ QEU\nMACData = U⎥⎥ V⎥⎥ QEU⎥⎥ QEV\nZ = MACTag\nCalculate the shared secret Z\nMACTag = MACMkeyMACData\nMACData = U⎥⎥ V⎥⎥ QEU⎥⎥ QEV \nFigure 13.4\nGeneration of shared secret. (From M. Khan, F. Amini, and J. Miˇsi´c, in\nMobile Ad-hoc and Sensor Networks, Springer, 2006. With permission.)\n13.5.3\nGeneration of Shared Secret\nBoth parties involved in the protocol will generate a shared secret based on\nunique identifiers (i.e., distinguished names for parties involved), symmetric\nmaster keys, and challenges received and owned by each party, as shown\nin Figure 13.4.\n1.\nEach party will generate a MACData by appending their identifiers\nand respective valid Challenges together as follows:\nMACData = U ||V||QEU||QEV\n2.\nEach party will calculate the MACTag (i.e., keyed hash) for M AC\nData using Mkey (master key for the device) as the key for keyed\nhash function as follows:\nMACTag = M AC MkeyMACData\n3.\nNow both parties involved have derived the same secret Z. (Note:\nThis is just a shared secret, not the link key. This shared secret will\nbe involved in deriving the link key, but is not the link key itself.)\nSet Z = MACTag\n" }, { "page_number": 438, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n427\n13.5.4\nDerivation of Link Key\nEach party involved will generate two cryptographic hashes (this is not the\nkeyed hash) of the shared secret as described in ANSI X9.63-2001 [1].\nHash1 = H(Z||01)\nHash2 = H(Z||02)\nThe hash value Hash2 will be the link key among two devices\n(Figure 13.5). Now for confirming that both parties have reached the same\nlink key (KeyData = Hash2), we will use value Hash1 as the key for gen-\nerating keyed hash values for confirming the stage of the protocol.\nMACKey = Hash1\n(13.1)\nKeyData = Hash2\n(13.2)\nK KeyData = Hash1||Hash2\n(13.3)\nCalculate the shared secret \nZ = MACTag \nDerive link key\nKeyData = H(Z⎥⎥ 02)\nand also\nMACKey = H(Z⎥⎥ 01)\nCalculate the shared secret \nZ = MACTag \nDerive link key\nKeyData = H(Z⎥⎥ 02)\nand also\nMACKey = H(Z⎥⎥ 01)\nInitiator \nU \nPAN \ncoordinator \nResponder \nV \nDevice \nU⎥⎥ V⎥⎥ QEU⎥⎥ QEV\nU⎥⎥ QEU\nBoth devices have generated link keys, now will\nconfirm that they have generated same link keys.\nFigure 13.5\nGeneration of link key.\n" }, { "page_number": 439, "text": "428\n■\nSecurity in Wireless Mesh Networks\n13.5.5\nConfirming Link Key\nUp to this stage of protocol, both parties are generating the same values\nand now they want to make sure that they have reached the same link key\nvalues, but they do not want to exchange the actual key at all. For this, they\nwill once again rely on keyed hash functions and now both devices will\ngenerate different MACTags based on different data values, but will use the\nsame key (i.e., MACKey) for generating the keyed hashes (MACTags).\n1.\nGeneration of MACTags: Initiator and responder devices will first\ngenerate MACData values and based on these values will generate\nMACTags. Initiator device D will receive the MACTag1 from the\nresponder device V and generate MACTag2 and send to device V.\nWe explain the generation of both MACData values and M AC\nT ags as follows. First, both devices will calculate MACData val-\nues:\nMACData1 = 0216||V||U ||QEU||QEV\nMACData2 = 0316||V||U ||QEU||QEV\nFrom the above MACData values both devices will generate the\nMACTags using the key M ACkey (Equation 13.1) as follows:\nMacT ag1 = M AC MacK eyMacData1\nMacT ag2 = M AC MacK eyMacData2\n2.\nConfirmation of MACTags: Now the initiator device D will receive\nMacT ag1 from the responder and responder device V will receive\nMACTag2 from device D and both will verify that the recieved\nMACTags are equal to corresponding calculated MACTags by each\ndevice. Now if this verification is successful, each device knows that\nthe other device has computed the correct link key, as shown in\nFigure 13.6.\n13.5.6\nCommunication Steps in SKKE Protocol\nSKKE protocol can be implemented in four major communication steps, as\ndescribed in ZigBee specification [4] as shown in Figure 13.7.\n1.\nSKKE-1: Initiator U will send the challenge QEU and wait for the\nchallenge QEV from responder V.\n" }, { "page_number": 440, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n429\nRecieve MACTag2 from U \ncalculate MACTag2 and compare \nwith received MACTag2, if valid \ngenerate ACK and send to U. \nGenerate MACTag1 and send to U \nMACTag1 = MACMacKeyMACData1 \nGenerate MACDatas\nMACTag1 \nACK \nMACTag2 \nMACData1 = 0216 V⎥⎥ U⎥⎥ QEU⎥⎥ QEV\nMACData2 = 0316 V⎥⎥ U⎥⎥ QEU⎥⎥ QEV\nMACData1 = 0216 V⎥⎥ U⎥⎥ QEU⎥⎥ QEV\nMACData2 = 0316 V⎥⎥ U⎥⎥ QEU⎥⎥ QEV\nBoth devices have confirmed that they have \nsame link keys \nResponder \nV \nInitiator \nU \nPAN \ncoordinator \nDevice \nU ⎜⎜V⎥⎥ QEU⎥⎥ QEV\nU⎥⎥ QEU\nGenerate MACDatas\nReceive MACTag1 from V\ncalculate MACTag1 and compare\nwith received MACTag1 and if valid\ngenerate MACTag2 and send to V.\nMACTag2 = MACMacKeyMACData2\nBoth devices have generated link keys, now will\nconfirm that they have generated same link keys\nFigure 13.6\nConfirmation of link keys.\n" }, { "page_number": 441, "text": "430\n■\nSecurity in Wireless Mesh Networks\nFigure 13.7\nCommunication steps in SKKE protocol.\n" }, { "page_number": 442, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n431\n2.\nSKKE-2: Responder V will receive the challenge QEU from initiator\nU , calculate its QEV , and in the same data packet will send the\nMacT ag1.\n3.\nSKKE-3: Initiator will verify the MacT ag1 and if it is verified suc-\ncessfully, will send its MacT ag2. Now the initiator has a link key,\nbut will wait for an acknowledgment that its MacT ag2 has been\nvalidated by the responder V.\n4.\nSKKE-4: Responder will receive and validate the MacT ag2 from the\ninitiator. If MacT ag2 validates successfully, the responder will send\nan acknowledgment and now both initiator and responder have\nlink keys. Once initiator receives this SKKE-4 message, keys estab-\nlishment is complete, and now regular secure communication can\nproceed using the link key among the initiator and the responder.\nAuthors have simulated the key exchange process in IEEE 802.15.4 on\ntop of the simulation model of this network and initial results confirm\nthe expected performance decrease of the overall network. They also have\nprovided data encryption by exchanging link keys between each device and\nclusterhead. The signature payload plays a big role on performance of the\ncluster. Also we have observed that the total access delay is higher when\nencryption and decryption are provided.\n13.6\nSummary\nIn this chapter we have outlined a number of problems in achieving the\ntarget of secure communication in wireless sensor networks. IEEE 802.15.4\ncluster-based wireless sensor network provides higher bandwidth links for\ninter-coordinator communication, and allows higher power resources at\nthe coordinator, but still the implementer of higher layers should make a\ngreat deal of effort in choosing the right keying model based on the appli-\ncation requirements. Even though ZigBee Alliance has outlined protocols\nregarding the key exchange, it is necessary to integrate them with the IEEE\n802.15.4 MAC protocol. Because key exchange protocols require downlink\ncommunications from the PAN coordinator to the ordinary nodes, it will\nconsume a lot of bandwidth. Therefore, a period of key exchange is a\ncrucial design parameter which has to match both security and bandwidth\nrequirements for the sensing application. Also, addition of the message au-\nthentication code at the end of the packet decreases the bandwidth which\nis left to the application and affects the complexity of the aggregation. We\nexpect that future work in this area (by us and other researchers) will de-\nliver the reasonable trade-off between the level of security and application\nbandwidth in large sensor networks implemented over interconnected IEEE\n802.15.4 clusters.\n" }, { "page_number": 443, "text": "432\n■\nSecurity in Wireless Mesh Networks\nReferences\n[1]\nANSI X9.63-2001, Public Key Cryptography for the Financial Services\nIndustry—Key Agreement and Key Transport Using Elliptic Curve Cryp-\ntography. American Bankers Association, 2001.\n[2]\nFIPS Pub. 198, The Keyed-Hash Message Authentication Code (HMAC).\nFederal Information Processing Standards Publication 198, U.S. Department\nof Commerce/N.I.S.T., 2002.\n[3]\nStandard for part 15.4: Wireless medium access control (MAC) and physical\nlayer (PHY) specifications for low rate wireless personal area networks\n(WPAN). IEEE Std 802.15.4, IEEE, 2003.\n[4]\nZ. Alliance.\nZigBee specification (ZigBee document 053474r06, version\n1.0), Dec. 2004.\n[5]\nR. Blom. An optimal class of symmetric key generation systems. In Proc.\nof the EUROCRYPT 84 Workshop on Advances in Cryptology: Theory and\nApplication of Cryptographic Techniques, pages 335–338, New York, 1985.\nSpringer-Verlag.\n[6]\nC. Blundo, A. D. Santis, A. Herzberg, S. Kutten, U. Vaccaro, and M. Yung.\nPerfectly-secure key distribution for dynamic conferences. In CRYPTO ’92:\nProceedings of the 12th Annual International Cryptology Conference on\nAdvances in Cryptology, pages 471–486, London, UK, 1993. Springer-Verlag.\n[7]\nM. Burmester and Y. Desmedt.\nA secure and efficient conference key\ndistribution system. In In Advances in Cryptology—EUROCRYPT 94, A. D.\nSantis, Ed., Lecture Notes in Computer Science, vol. 950, pages 275–286,\nNew York, 1994. Springer-Verlag.\n[8]\nH. Chan, A. Perrig, and D. Song. Random key predistribution schemes\nfor sensor networks. In SP ’03: Proceedings of the 2003 IEEE Symposium\non Security and Privacy, page 197, Washington, DC, 2003. IEEE Computer\nSociety.\n[9]\nW. Du, J. Deng, Y. S. Han, and P. K. Varshney.\nA pairwise key pre-\ndistribution scheme for wireless sensor networks. In CCS ’03: Proceedings\nof the 10th ACM Conference on Computer and Communications Security,\npages 42–51, New York, 2003. ACM Press.\n[10]\nL. Eschenauer and V. D. Gligor. A key-management scheme for distributed\nsensor networks. In CCS ’02: Proceedings of the 9th ACM Conference on\nComputer and Communications Security, pages 41–47, New York, 2002.\nACM Press.\n[11]\nV. Gupta, M. Millard, S. Fung, Y. Zhu, N. Gura, H. Eberle, and S. C. Shantz.\nSizzle: A standards-based end-to-end security architecture for the embed-\nded Internet (best paper). In PERCOM ’05: Proceedings of the Third IEEE\nInternational Conference on Pervasive Computing and Communications,\npages 247–256, Washington, DC, 2005. IEEE Computer Society.\n[12]\nH. Kang and X. Li. Power-aware sensor selection in wireless sensor net-\nworks. Scalable Software Systems Laboratory, Computer Science Depart-\nment, Oklahoma State University, 2006.\n[13]\nD. Liu, P. Ning, and R. Li. Establishing pairwise keys in distributed sensor\nnetworks. ACM Trans. Inf. Syst. Secur., 8(1):41–77, 2005.\n" }, { "page_number": 444, "text": "Security in IEEE 802.15.4 Cluster-Based Networks\n■\n433\n[14]\nA. Menezes, P. van Oorschot, and S. Vanstone. Handbook of Applied Cryp-\ntography. Boca Raton, FL, 1997, CRC Press.\n[15]\nJ. Miˇsi´c, S. Shafi, and V. B. Miˇsi´c. Performance of beacon enabled IEEE\n802.15.4 cluster with downlink and uplink tarffic. IEEE Transactions on\nParallel and Distributed Systems, 17(4):361–377, Apr. 2006.\n[16]\nN. Sastry and D. Wagner. Security considerations for IEEE 802.15.4 net-\nworks. In WiSe ’04: Proceedings of the 2004 ACM Workshop on Wireless\nSecurity, pages 32–42, New York, 2004. ACM Press.\n[17]\nG. W. M. Steiner, and M. Tsudik. Key agreement in dynamic peer groups.\nIn IEEE Transactions on Parallel and Distributed Systems, pages 769–780,\nWashington, DC, Aug. 2000. IEEE Computer Society.\n[18]\nA. S. Wander, N. Gura, H. Eberle, V. Gupta, and S. C. Shantz. Energy anal-\nysis of public-key cryptography for wireless sensor networks. In PERCOM\n’05: Proceedings of the Third IEEE International Conference on Pervasive\nComputing and Communications, pages 324–328, Washington, DC, 2005.\nIEEE Computer Society.\n[19]\nS. Zhu, S. Setia, and S. Jajodia. LEAP: efficient security mechanisms for\nlarge-scale distributed sensor networks.\nIn CCS ’03: Proceedings of the\n10th ACM Conference on Computer and Communications Security, pages\n62–72, New York, 2003. ACM Press.\n[20]\nM. Khan, F. Amini, and J. Miˇsi´c. Key exchange in 802.15.4 networks and its\nimplications. In Mobile Ad-hoc and Sensor Networks, H. Zhang, S. Olariu,\nJ. Cao, and D. B. Johnson (Eds.), (pp. 497–508). Berlin, 2006. Springer.\n" }, { "page_number": 445, "text": "" }, { "page_number": 446, "text": "Chapter 14\nSecurity in Wireless\nSensor Networks\nYong Wang, Garhan Attebury, and Byrav Ramamurthy\nContents\n14.1\nIntroduction ........................................................ 437\n14.2\nBackground......................................................... 439\n14.2.1\nCommunication Architecture.............................. 439\n14.2.2\nConstraints in WSNs ....................................... 440\n14.2.2.1 Energy ........................................... 441\n14.2.2.2 Computation ..................................... 441\n14.2.2.3 Memory .......................................... 441\n14.2.2.4 Transmission Range ............................. 441\n14.2.3\nSecurity Requirements..................................... 443\n14.2.4\nThreat Model .............................................. 443\n14.2.5\nEvaluation ................................................. 444\n14.3\nAttacks in Sensor Networks ........................................ 444\n14.3.1\nPhysical Layer ............................................. 445\n14.3.1.1 Jamming ......................................... 445\n14.3.1.2 Tampering ....................................... 446\n14.3.2\nLink Layer.................................................. 446\n14.3.2.1 Collisions ........................................ 446\n14.3.2.2 Exhaustion ....................................... 447\n14.3.2.3 Unfairness ....................................... 447\n14.3.3\nNetwork and Routing Layer............................... 447\n14.3.3.1 Spoofed, Altered, or Replayed\nRouting Information ............................. 447\n435\n" }, { "page_number": 447, "text": "436\n■\nSecurity in Wireless Mesh Networks\n14.3.3.2 Selective Forwarding ............................ 448\n14.3.3.3 Sinkhole ......................................... 448\n14.3.3.4 Sybil .............................................. 448\n14.3.3.5 Wormholes....................................... 449\n14.3.3.6 Hello Flood Attacks ............................. 449\n14.3.3.7 Acknowledgment Spoofing ..................... 449\n14.3.4\nTransport Layer ............................................ 449\n14.3.4.1 Flooding ......................................... 450\n14.3.4.2 Desynchronization............................... 450\n14.4\nCryptography in WSNs ............................................. 451\n14.4.1\nPublic Key Cryptography in WSNs ....................... 451\n14.4.2\nSymmetric Key Cryptography in WSNs ................... 455\n14.4.3\nOpen Research Issues ..................................... 456\n14.5\nKey Management Protocols ........................................ 456\n14.5.1\nNetwork Structure-Based Key Management Protocols ... 457\n14.5.1.1 Centralized Key Management Schemes......... 458\n14.5.1.2 Distributed Key Management Schemes ......... 458\n14.5.2\nKey Management Protocols Based on the Probability\nof Key Sharing............................................. 458\n14.5.2.1 Deterministic Approaches ....................... 458\n14.5.2.2 Probabilistic Approaches ........................ 460\n14.5.3\nOpen Research Issues ..................................... 463\n14.6\nSecure Routing Protocols .......................................... 465\n14.6.1\nBroadcast Authentication.................................. 467\n14.6.2\nSecure Routing ............................................ 469\n14.6.3\nOpen Research Issues ..................................... 473\n14.7\nSecure Data Aggregation ........................................... 473\n14.7.1\nPlaintext-Based Secure Data Aggregation ................ 474\n14.7.2\nCiphertext-Based Secure Data Aggregation .............. 477\n14.7.3\nOpen Research Issues ..................................... 478\n14.8\nIntrusion Detection ................................................. 478\n14.8.1\nIntrusion Detection in WSNs.............................. 479\n14.8.2\nOpen Research Issues ..................................... 481\n14.9\nSecurity in WSNs: Future Directions ............................... 482\n14.10 Summary............................................................ 483\nAcknowledgments ........................................................ 483\nReferences................................................................. 483\nWireless sensor networks (WSNs) are used in many applications in military,\necological, and health-related areas. These applications often include the\nmonitoring of sensitive information such as enemy movement on the battle-\nfield or the location of personnel in a building. Security is therefore impor-\ntant in WSNs. However, WSNs suffer from many constraints including low\n" }, { "page_number": 448, "text": "Security in Wireless Sensor Networks\n■\n437\ncomputation capability, small memory, limited energy resources, suscep-\ntibility to physical capture, and the use of insecure wireless communica-\ntion channels. These constraints make security challenging in WSNs. In this\nchapter, we present a survey of security issues in WSNs. First we outline the\nconstraints, security requirements, and attacks with corresponding counter-\nmeasures in WSNs. We then present a holistic view of security issues. These\nissues are classified into five categories: cryptography, key management,\nsecure routing, secure data aggregation, and intrusion detection. Along the\nway we highlight advantages and disadvantages of various WSN security\nprotocols and further compare and evaluate these protocols based on each\nof these five categories. We also point out the open research issues in each\nsub-area and conclude with possible future research directions on security\nin WSNs.\n14.1\nIntroduction\nAdvances in wireless communication and electronics have enabled the de-\nvelopment of low-cost, low-power, multi-functional sensor nodes. These\ntiny sensor nodes, consisting of sensing, data processing, and communi-\ncation components, make it possible to deploy WSNs, which represent a\nsignificant improvement over traditional wired sensor networks. WSNs can\ngreatly simplify system design and operation as the environment being\nmonitored does not require the communication or energy infrastructure\nassociated with wired networks [1].\nWSNs are expected to be solutions to many applications, such as detect-\ning and tracking the passage of troops and tanks on a battlefield, monitoring\nenvironmental pollutants, measuring traffic flows on roads, and tracking the\nlocation of personnel in a building. Many sensor networks have mission-\ncritical tasks and thus require security to be considered [2,3]. Improper use\nof information, or using forged information, may cause unwanted informa-\ntion leakage and provide inaccurate results.\nWhile some aspects of WSNs are similar to traditional wireless ad hoc\nnetworks, important distinctions exist which greatly affect how security is\nachieved. The differences between sensor networks and ad hoc networks\n[4] are:\n■\nThe number of sensor nodes in a sensor network can be several\norders of magnitude higher than the nodes in an ad hoc network.\n■\nSensor nodes are densely deployed.\n■\nSensor nodes are prone to failures due to harsh environments and\nenergy constraints.\n■\nThe topology of a sensor network changes very frequently due to\nfailures or mobility.\n" }, { "page_number": 449, "text": "438\n■\nSecurity in Wireless Mesh Networks\n■\nSensor nodes are limited in computation, memory, and power\nresources.\n■\nSensor nodes may not have global identification.\nThese differences greatly affect how secure data transfer schemes are imple-\nmented in WSNs. For example, the use of radio transmission, along with the\nconstraints of small size, low cost, and limited energy, make WSNs more\nsusceptible to denial-of-service attacks [5]. Advanced anti-jamming tech-\nniques such as frequency-hopping spread spectrum and physical tamper-\nproofing of nodes are generally impossible in a sensor network due to the\nrequirements of greater design complexity and higher energy consump-\ntion [5]. Furthermore, the limited energy and processing power of nodes\nmakes the use of public key cryptography nearly impossible. While the re-\nsults from recent studies show that public key cryptography might be feasi-\nble in sensor networks [6,7], it remains for the most part infeasible in WSNs.\nInstead, most security schemes make use of symmetric key cryptography.\nOne thing required in either case is the use of keys for secure communica-\ntion. Managing key distribution is not unique to WSNs, but again constraints\nsuch as small memory capacity make centralized keying techniques impos-\nsible. Straight pairwise key sharing between every two nodes in a network\ndoes not scale to large networks with tens of thousands of nodes as the\nstorage requirements are too high. A security scheme in WSNs must provide\nefficient key distribution while maintaining the ability for communication\nbetween all relevant nodes.\nIn addition to key distribution, secure routing protocols must be con-\nsidered. These protocols are concerned with how a node sends messages\nto other nodes or a base station. A key challenge is that of authenticated\nbroadcast. Existing authenticated broadcast methods often rely on public\nkey cryptography and include high computational overhead, making them\ninfeasible in WSNs. Secure routing protocols proposed for use in WSNs,\nsuch as security protocols for sensor networks (SPINS) [8], must consider\nthese factors. Additionally, the constraint on energy in WSNs leads to the\ndesire for data aggregation. This aggregation of sensor data needs to be\nsecure to ensure information integrity and confidentiality [9,10]. Although\nthis is achievable through cryptography, an aggregation scheme must take\ninto account the constraints in WSNs and the unique characteristics of the\ncryptography and routing schemes. It is also desirable for secure data ag-\ngregation protocols to be flexible, allowing lower levels of security for less\nimportant data, saving energy, and higher levels of security for more sen-\nsitive data, consuming more energy.\nAs with any network, the awareness of compromised nodes and attacks\nis desirable. Many security schemes provide assurance that data remains\nintact and communication unaffected as long as fewer than t nodes are\ncompromised [11]. The ability of a node or base station to detect when\n" }, { "page_number": 450, "text": "Security in Wireless Sensor Networks\n■\n439\nother nodes are compromised enables them to take action, either ignoring\nthe compromised data or reconfiguring the network to eliminate the threat.\nThe remainder of this chapter discusses the above areas in more detail\nand considers how they are all required to form a complete WSN security\nscheme. A few existing surveys on security issues in ad hoc networks can\nbe found in [12–14]. However, only small sections of these surveys focus on\nWSNs. A recent survey paper on security issues in mobile ad hoc networks\nalso included an overview of security issues in WSNs [15]. However, the\npaper did not discuss cryptography and intrusion detection issues. Further,\nit included only a small portion of the available literature on security in\nWSNs.\nThe rest of the chapter is organized as follows. Section 14.2 presents\nbackground information on WSNs. Section 14.3 discusses attacks in the dif-\nferent network layers of sensor networks, followed by Section 14.4, which\nfocuses on the selection of cryptography in WSNs. Section 14.5 focuses on\nkey management, Section 14.6 on secure routing schemes, Section 14.7 on\nsecure data aggregation, and Section 14.8 on intrusion detection systems.\nSection 14.9 discusses future research directions on security in WSNs, and\nSection 14.10 concludes the chapter.\n14.2\nBackground\n14.2.1\nCommunication Architecture\nA WSN is usually composed of hundreds or thousands of sensor nodes.\nThese sensor nodes are often densely deployed in a sensor field and\nhave the capability to collect data and route data back to a base sta-\ntion (BS). A sensor consists of four basic parts: a sensing unit, a pro-\ncessing unit, a transceiver unit, and a power unit [4]. It may also have\nadditional application-dependent components such as a location finding\nsystem, power generator, and mobilizer (Figure 14.1). Sensing units are\nusually composed of two sub-units: sensors and analog-to-digital convert-\ners (ADCs). The ADCs convert the analog signals produced by the sensors\nto digital signals based on the observed phenomenon. The processing unit,\nwhich is generally associated with a small storage unit, manages the pro-\ncedures that make the sensor node collaborate with the other nodes. A\ntransceiver unit connects the node to the network. One of the most impor-\ntant units is the power unit. A power unit may be finite, such as a single\nbattery, or may be supported by power scavenging devices, such as solar\ncells. Most of the sensor network routing techniques and sensing tasks\nrequire knowledge of location, which is provided by a location finding\nsystem. Finally, a mobilizer may sometimes be needed to move the sensor\nnode depending on the application.\n" }, { "page_number": 451, "text": "440\n■\nSecurity in Wireless Mesh Networks\nPosition finding system\nMobilizer\nPower unit\nPower\ngenerator\nBS\nSensor node\nSensor\nADC\nSensing unit\nProcessor\nStorage\nTransceiver\nTransmission\nunit\nProcessing unit\nInternet\nUser\nFigure 14.1\nThe components of a sensor node. (From Y. Wang, G. Attebury, and\nB. Ramamurthy, IEEE Communications Surveys and Tutorials, Vol. 8, no. 2, pp. 2–23,\n2006. With permission.)\nThe protocol stack used in sensor nodes contains physical, data link,\nnetwork, transport, and application layers defined as follows [4]:\n■\nPhysical layer: Responsible for frequency selection, carrier frequency\ngeneration, signal deflection, modulation, and data encryption.\n■\nData link layer: Responsible for the multiplexing of datastreams,\ndata frame detection, medium access, and error control. This layer\nensures reliable point-to-point and point-to-multipoint connections.\n■\nNetwork layer: Responsible for specifying the assignment of ad-\ndresses and how packets are forwarded.\n■\nTransport layer: Responsible for specifying how the reliable trans-\nport of packets will take place.\n■\nApplication layer: Responsible for specifying how the data is re-\nquested and provided for both the individual sensor nodes and the\ninteractions with the end user.\n14.2.2\nConstraints in WSNs\nIndividual sensor nodes in a WSN are inherently resource constrained. They\nhave limited processing capability, storage capacity, and communication\n" }, { "page_number": 452, "text": "Security in Wireless Sensor Networks\n■\n441\nbandwidth. Each of these limitations is due in part to the two greatest con-\nstraints: limited energy and physical size. Table 14.1 shows several currently\navailable sensor node platforms. The design of security services in WSNs\nmust consider the hardware constraints of the sensor nodes.\n14.2.2.1\nEnergy\nEnergy consumption in sensor nodes can be categorized into three parts:\n1.\nEnergy for the sensor transducer\n2.\nEnergy for communication among sensor nodes\n3.\nEnergy for microprocessor computation\nThe study in [20,21] found that each bit transmitted in WSNs consumes\nabout as much power as executing 800 to 1000 instructions. Thus, commu-\nnication is more costly than computation in WSNs. Any message expansion\ncaused by security mechanisms comes at a significant cost. Further, higher\nsecurity levels in WSNs usually correspond to more energy consumption\nfor cryptographic functions. Thus, WSNs could be divided into different\nsecurity levels depending on energy cost [22,23].\n14.2.2.2\nComputation\nThe embedded processors in sensor nodes are generally not as powerful\nas those in nodes of a wired or ad hoc network. As such, complex crypto-\ngraphic algorithms cannot be used in WSNs.\n14.2.2.3\nMemory\nMemory in a sensor node usually includes flash memory and RAM. Flash\nmemory is used for storing downloaded application code and RAM is used\nfor storing application programs, sensor data, and intermediate computa-\ntions. There is usually not enough space to run complicated algorithms after\nloading OS and application code. In the SmartDust project, for example,\nTinyOS consumes about 3500 bytes of instruction memory, leaving only\n4500 bytes for security and applications [20,21]. This makes it impractical\nto use the majority of current security algorithms [8]. With an Intel Mote, the\nsituation is slightly improved, but still far from meeting the requirements\nof many algorithms.\n14.2.2.4\nTransmission Range\nThe communication range of sensor nodes is limited both technically and\nby the need to conserve energy. The actual range achieved from a given\ntransmission signal strength is dependent on various environmental factors\nsuch as weather and terrain.\n" }, { "page_number": 453, "text": "442\n■\nSecurity in Wireless Mesh Networks\nTable 14.1\nVariety of Real-Life Sensor Nodes\nBerkeley Mote [16]\nWeC\nrene2\nrene2\ndot\nmica\nEYES [17]\nMedusa MK-2 [18]\nImote [19]\nMonth/year\n09/99\n10/00\n06/01\n08/01\n02/02\n03/02\n09/02\n01/03\nCPU\nAT90LS8535\nATmega163\nATmega103a\nMSP 430F149\n40 MHz ARM THUMB\nARM core 12 MHz\nProg. mem.\n8 kb\n16 kb\n128 kb\n60 kb\n1 MB\n512 kb\nRAM\n0.5 kb\n1 kb\n4 kb\n2 kb\n136 kb\n64 kb\nRadio\nTR1000 916 MHz\nTR1001 868.35 MHz\nTR1000 916 MHz\nBT 2.4 GHz\nRate\n10 kbps\n10/40 kbps\n115 kbps\n115 kbps\n100 kbps\naLater versions are an ATmega128 running in 103 mode.\nSource: Y. Wang, G. Attebury, and B. Ramamurthy, IEEE Communications Surveys and Tutorials, Vol. 8, no. 2, pp. 2–23, 2006. With permission.\n" }, { "page_number": 454, "text": "Security in Wireless Sensor Networks\n■\n443\n14.2.3\nSecurity Requirements\nThe goal of security services in WSNs is to protect the information and\nresources from attacks and misbehavior. The security requirements in WSNs\ninclude:\n■\nAvailability: Ensures that the desired network services are available\neven in the presence of denial-of-service attacks.\n■\nAuthorization: Ensures that only authorized sensors can be involved\nin providing information to network services.\n■\nAuthentication: Ensures that the communicating node is the one that\nit claims to be.\n■\nConfidentiality: Ensures that a given message cannot be understood\nby anyone other than the desired recipients.\n■\nIntegrity: Ensures that a message sent from one node to another is\nnot altered by unauthorized or unknown means.\n■\nNon-repudiation: Denotes that a node cannot deny sending a mes-\nsage it has previously sent.\n■\nFreshness: Implies that the data is recent and ensures that no adver-\nsary can replay old messages.\nThe security services in WSNs are usually centered around cryptogra-\nphy. However, because of the constraints in WSNs, many already-existing\nsecure algorithms are not practical for use. We discuss this problem in\nSection 14.4.\n14.2.4\nThreat Model\nIn WSNs, it is usually assumed that an attacker may know the security\nmechanisms that are deployed in a sensor network, and may be able to\ncompromise a node or even physically capture a node. Due to the high cost\nof deploying tamper-resistant sensor nodes, most WSN nodes are viewed\nas non-tamper-resistant. Further, once a node is compromised, the attacker\nis capable of stealing the key materials contained within that node.\nBase stations in WSNs are usually regarded as trustworthy. Most research\nstudies focus on secure routing between sensors and the base station. Deng\net al. considered strategies against threats which can lead to the failure of\nthe base station [24].\nAttacks in sensor networks can be classified into the following cate-\ngories:\n■\nOutsider versus insider attacks: Outside attacks are defined as at-\ntacks from nodes which do not belong to a WSN. Inside attacks\noccur when legitimate nodes of a WSN behave in unintended or\nunauthorized ways.\n" }, { "page_number": 455, "text": "444\n■\nSecurity in Wireless Mesh Networks\n■\nPassive versus active attacks: Passive attacks include eavesdropping\non packets exchanged within a WSN. Active attacks involve some\nmodifications of the datastream or the creation of a false stream.\n■\nMote-class versus laptop-class attacks: In mote-class attacks, an ad-\nversary attacks a WSN by using a few nodes with similar capabilities\nto the network nodes. In laptop-class attacks, an adversary can use\nmore powerful devices such as a laptop to attack a WSN. These de-\nvices have greater transmission range, processing power, and energy\nreserves than the network nodes.\n14.2.5\nEvaluation\nWe suggest using the following metrics to evaluate whether a security\nscheme is appropriate in WSNs.\n■\nSecurity: A security scheme has to meet the requirements discussed\nin Section 14.2.3.\n■\nResiliency: In case a few nodes are compromised, a security scheme\ncan still protect against the attacks.\n■\nEnergy efficiency: A security scheme must be energy-efficient to\nmaximize node and network lifetime.\n■\nFlexibility: The key management needs to be flexible to allow for\ndifferent network deployment methods such as random node scat-\ntering and pre-determined node placement.\n■\nScalability: A security scheme should be able to scale without com-\npromising the security requirements.\n■\nFault-tolerance: A security scheme should continue to provide secu-\nrity services in the presence of faults such as failed nodes.\n■\nSelf-healing: Sensors may fail or run out of energy. The remain-\ning sensors may need to be re-organized to maintain a set level of\nsecurity.\n■\nAssurance: Assurance is the ability to disseminate different informa-\ntion at different levels to end users [25]. A security scheme should\noffer choices as to desired reliability, latency, and so on.\n14.3\nAttacks in Sensor Networks\nWSNs are vulnerable to various types of attacks. According to the security\nrequirements in WSNs, these attacks can be categorized [3] as:\n■\nAttacks on secrecy and authentication: Standard cryptographic tech-\nniques can protect the secrecy and authenticity of communication\n" }, { "page_number": 456, "text": "Security in Wireless Sensor Networks\n■\n445\nchannels from outsider attacks such as eavesdropping, packet replay\nattacks, and modification or spoofing of packets.\n■\nAttacks on network availability: Attacks on availability are often re-\nferred to as denial-of-service (DoS) attacks. DoS attacks may target\nany layer of a sensor network.\n■\nStealthy attacks against service integrity: In a stealthy attack, the goal\nof the attacker is to make the network accept a false data value. For\nexample, an attacker compromises a sensor node and injects a false\ndata value through that sensor node.\nIn these attacks, keeping the sensor network available for its intended\nuse is essential. DoS attacks against WSNs may permit real-world damage\nto the health and safety of people [5]. In this section, we focus only on\nDoS attacks and their countermeasures in sensor networks. Section 14.6\ndiscusses attacks on secrecy and authentication and Section 14.8 discusses\nstealthy attacks and countermeasures.\nThe DoS attack usually refers to an adversary’s attempt to disrupt, sub-\nvert, or destroy a network. However, a DoS attack can be any event that\ndiminishes or eliminates a network’s capacity to perform its expected func-\ntion [5]. Sensor networks are usually divided into layers, and this layered\narchitecture makes WSNs vulnerable to DoS attacks, which may occur in\nany layer of a sensor network.\nPrevious discussions on DoS attacks in WSNs can be found in [3,5,26,27].\nThe remainder of this section summarizes possible DoS attacks and coun-\ntermeasures in each layer of a sensor network.\n14.3.1\nPhysical Layer\nThe physical layer is responsible for frequency selection, carrier frequency\ngeneration, signal detection, modulation, and data encryption [4]. As with\nany radio-based medium there exists the possibility of jamming in WSNs. In\naddition, nodes in WSNs may be deployed in hostile or insecure environ-\nments where an attacker has easy physical access. These two vulnerabilities\nare explored in this subsection.\n14.3.1.1\nJamming\nJamming is a type of attack which interferes with the radio frequencies\nthat a network’s nodes are using [3,5]. A jamming source may either be\npowerful enough to disrupt the entire network or less powerful and only\nable to disrupt a smaller portion of the network. Even with lesser-powered\njamming sources, such as a small compromised subset of the network’s\nsensor nodes, an adversary has the potential to disrupt the entire network\nprovided the jamming sources are randomly distributed in the network.\n" }, { "page_number": 457, "text": "446\n■\nSecurity in Wireless Mesh Networks\nTypical defenses against jamming involve variations of spread-spectrum\ncommunication such as frequency hopping and code spreading [5].\nFrequency-hopping spread spectrum (FHSS) is a method of transmitting\nsignals by rapidly switching a carrier among many frequency channels us-\ning a pseudo-random sequence known to both transmitter and receiver.\nWithout being able to follow the frequency selection sequence, an attacker\nis unable to jam the frequency being used at a given moment in time.\nHowever, as the range of possible frequencies is limited, an attacker may\ninstead jam a wide section of the frequency band.\nCode spreading is another technique used to defend against jamming\nattacks and is common in mobile networks. However, this technique re-\nquires greater design complexity and energy restricting its use in WSNs. In\ngeneral, to maintain low cost and low power requirements, sensor devices\nare limited to single-frequency use and are therefore highly susceptible to\njamming attacks.\n14.3.1.2\nTampering\nAnother physical layer attack is tampering [5]. Given physical access to a\nnode, an attacker can extract sensitive information such as cryptographic\nkeys or other data on the node. The node may also be altered or re-\nplaced to create a compromised node which the attacker controls. One\ndefense against this attack involves tamper-proofing the node’s physical\npackage [5]. However, it is usually assumed that the sensor nodes are not\ntamper-proofed in WSNs due to the additional cost. This indicates that a\nsecurity scheme must consider the situation in which sensor nodes are\ncompromised.\n14.3.2\nLink Layer\nThe data link layer is responsible for the multiplexing of datastreams, data\nframe detection, medium access, and error control [4]. It ensures reliable\npoint-to-point and point-to-multipoint connections in a communications\nnetwork. Attacks at the link layer include purposely introduced collisions,\nresource exhaustion, and unfairness. This sub-section looks at each of these\nlink layer attack categories [5].\n14.3.2.1\nCollisions\nA collision occurs when two nodes attempt to transmit on the same fre-\nquency simultaneously [5]. When packets collide, a change will likely occur\nin the data portion causing a checksum mismatch at the receiving end. The\npacket will then be discarded as invalid. An adversary may strategically\ncause collisions in specific packets such as ACK control messages. A pos-\nsible result of such collisions is the costly exponential back-off in certain\nmedia access control protocols.\n" }, { "page_number": 458, "text": "Security in Wireless Sensor Networks\n■\n447\nA typical defense against collisions is the use of error-correcting codes [5].\nMost codes work best with low levels of collisions such as those caused\nby environmental or probabilistic errors. However, these codes also add\nadditional processing and communication overhead. It is reasonable to as-\nsume that an attacker will always be able to corrupt more than what can\nbe corrected. Although it is possible to detect these malicious collisions, no\ncomplete defenses against them are known at this time.\n14.3.2.2\nExhaustion\nRepeated collisions can also be used by an attacker to cause resource\nexhaustion [5]. For example, a naive link layer implementation may contin-\nuously attempt to retransmit the corrupted packets. Unless these hopeless\nretransmissions are discovered or prevented, the energy reserves of the\ntransmitting node and those surrounding it will be quickly depleted.\nA possible solution is to apply rate limits to the admission control in the\nmedium access control protocol such that the network can ignore excessive\nrequests preventing the energy drain caused by repeated transmissions [5].\nA second technique is to use time-division multiplexing where each node\nis allotted a time slot in which it can transmit [5]. This eliminates the need\nof arbitration for each frame and can solve the indefinite postponement\nproblem in a back-off algorithm. However, it is still susceptible to collisions.\n14.3.2.3\nUnfairness\nUnfairness can be considered a weak form of a DoS attack [5]. An attacker\nmay cause unfairness in a network by intermittently using the above link\nlayer attacks. Instead of outright preventing access to a service, an attacker\ncan degrade it to give them an advantage such as causing other nodes in a\nreal-time medium access control protocol to miss their transmission dead-\nline. The use of small frames lessens the effect of such attacks by reducing\nthe amount of time an attacker can capture the communication channel.\nHowever, this technique often reduces efficiency and is susceptible to fur-\nther unfairness such as an attacker trying to retransmit quickly instead of\nrandomly delaying.\n14.3.3\nNetwork and Routing Layer\nThe network and routing layer of sensor networks is usually designed ac-\ncording to the following principles [4]:\n■\nPower efficiency is an important consideration.\n■\nSensor networks are mostly data-centric.\n■\nAn ideal sensor network has attribute-based addressing and location\nawareness.\nThe attacks in network and routing layer are discussed next:\n" }, { "page_number": 459, "text": "448\n■\nSecurity in Wireless Mesh Networks\n14.3.3.1\nSpoofed, Altered, or Replayed Routing Information\nThe most direct attack against a routing protocol in any network is to\ntarget the routing information itself as it is exchanged between nodes. An\nattacker may spoof, alter, or replay routing information to disrupt traffic in\nthe network [26]. These disruptions include the creation of routing loops,\nattracting or repelling network traffic from select nodes, extending and\nshortening source routes, generating fake error messages, partitioning the\nnetwork, and increasing end-to-end latency.\nA countermeasure against spoofing and alteration is to append a MAC\nafter the message. By adding a MAC to the message, the receivers can\nverify whether the messages have been spoofed or altered. To defend\nagainst replayed information, counters or timestamps can be included in the\nmessages [8].\n14.3.3.2\nSelective Forwarding\nA significant assumption made in multi-hop networks is that all nodes in\nthe network will accurately forward received messages. An attacker may\ncreate malicious nodes which selectively forward only certain messages\nand simply drop others [26]. A specific form of this attack is the black hole\nattack in which a node drops all messages it receives. One defense against\nselective forwarding attacks is using multiple paths to send data [26]. A\nsecond defense is to detect the malicious node or assume it has failed and\nseek an alternative route.\n14.3.3.3\nSinkhole\nIn a sinkhole attack, an attacker makes a compromised node look more\nattractive to surrounding nodes by forging routing information [5,26]. The\nend result is that surrounding nodes will choose the compromised node as\nthe next node to route their data through. This type of attack makes selective\nforwarding very simple as all traffic from a large area in the network will\nflow through the adversary’s node.\n14.3.3.4\nSybil\nThe Sybil attack is a case where one node presents more than one identity\nto the network [3,26,27]. Protocols and algorithms which are easily affected\ninclude fault-tolerant schemes, distributed storage, and network topology\nmaintenance. For example, a distributed storage scheme may rely on there\nbeing three replicas of the same data to achieve a given level of redun-\ndancy. If a compromised node pretends to be two of the three nodes, the\n" }, { "page_number": 460, "text": "Security in Wireless Sensor Networks\n■\n449\nalgorithms used may conclude that redundancy has been achieved although\nin reality it has not.\n14.3.3.5\nWormholes\nA wormhole is a low latency link between two portions of the network over\nwhich an attacker replays network messages [26]. This link may be estab-\nlished either by a single node forwarding messages between two adjacent\nbut otherwise non-neighboring nodes or by a pair of nodes in different parts\nof the network communicating with each other. The latter case is closely\nrelated to the sinkhole attack as an attacking node near the base station can\nprovide a one-hop link to that base station via the other attacking node in a\ndistant part of the network. Hu et al. presented a novel and general mech-\nanism called packet leashes for detecting and defending against wormhole\nattacks [28]. Two types of leashes were introduced: geographic leashes and\ntemporal leashes. The proposed mechanisms can also be used in WSNs.\n14.3.3.6\nHello Flood Attacks\nMany protocols which use Hello packets make the naive assumption that\nreceiving such a packet means the sender is within radio range and is there-\nfore a neighbor. An attacker may use a high-powered transmitter to trick a\nlarge area of nodes into believing they are neighbors of that transmitting\nnode [26]. If the attacker falsely broadcasts a superior route to the base\nstation, all of these nodes will attempt transmitting to the attacking node\ndespite many being out of radio range in reality.\n14.3.3.7\nAcknowledgment Spoofing\nRouting algorithms used in sensor networks sometimes require acknowl-\nedgments to be used. An attacking node can spoof the acknowledgments of\noverheard packets destined for neighboring nodes to provide false informa-\ntion to those neighboring nodes [26]. An example of such false information\nis claiming that a node is alive when in fact it is dead.\n14.3.4\nTransport Layer\nThe transport layer is responsible for managing end-to-end connections [4].\nTwo possible attacks in this layer, flooding and desynchronization, are\ndiscussed in this sub-section.\n" }, { "page_number": 461, "text": "450\n■\nSecurity in Wireless Mesh Networks\n14.3.4.1\nFlooding\nWhenever a protocol is required to maintain state at either end of a connec-\ntion it becomes vulnerable to memory exhaustion through flooding [5]. An\nattacker may repeatedly make new connection requests until the resources\nrequired by each connection are exhausted or reach a maximum limit. In\neither case, further legitimate requests will be ignored. One proposed solu-\ntion to this problem is to require that each connecting client demonstrates\nits commitment to the connection by solving a puzzle [5]. The idea is that a\nconnecting client will not needlessly waste its resources creating unneces-\nsary connections. Given an attacker does not likely have infinite resources,\nit will be impossible for him to create new connections fast enough to\ncause resource starvation on the serving node. Although these puzzles do\ninclude processing overhead, this technique is more desirable than exces-\nsive communication.\n14.3.4.2\nDesynchronization\nDesynchronization refers to the disruption of an existing connection [5].\nAn attacker may, for example, repeatedly spoof messages to an end host\ncausing that host to request the retransmission of missed frames. If timed\ncorrectly, an attacker may degrade or even prevent the ability of the end\nhosts to successfully exchange data causing them instead to waste energy\nattempting to recover from errors which never really existed.\nA possible solution to this type of attack is to require authentication of all\npackets communicated between hosts [5]. Provided that the authentication\nmethod is itself secure, an attacker will be unable to send the spoofed\nmessages to the end hosts.\nTable 14.2 shows the possible DoS attacks and countermeasures in\nWSNs.\nIn the following sections we discuss cryptography, key management\nprotocols, secure routing protocols, secure data aggregation, and intrusion\ndetection for WSNs. For the remainder of this article we use the following\nnotation:\n■\nA, B are principals such as communicating nodes.\n■\nIDA denotes the sensor identifier of node A.\n■\nNA is a nonce generated by A (a nonce is an unpredictable bit string,\nusually used to achieve freshness).\n■\nKAB denotes the secret pairwise key shared between A and B.\n■\nMK is the encryption of message M with key K .\n■\nMAC(K, M) denotes the computation of the message authentication\ncode of message M with key K .\n■\nA −→B denotes A unicasts a message to B.\n■\nA −→∗denotes A broadcasts a message to its neighbors.\n" }, { "page_number": 462, "text": "Security in Wireless Sensor Networks\n■\n451\nTable 14.2\nSensor Network Layers and Denial-of-Service Defenses\nNetwork layer\nAttacks\nDefense\nPhysical\nJamming\nSpread-spectrum, priority messages,\nlower duty cycle, region mapping,\nmode change\nTampering\nTamper-proofing, hiding\nLink\nCollision\nError-correcting code\nExhaustion\nRate limitation\nUnfairness\nSmall frames\nNetwork\nSpoofed, altered, or\nEgress filtering, authentication,\nand routing\nreplayed routing\nmonitoring\ninformation\nSelective forwarding\nRedundancy, probing\nSinkhole\nAuthentication, monitoring, redundancy\nSybil\nAuthentication, probing\nWormholes\nAuthentication, packet leashes by\nusing geographic and temporal information\nHello flood attacks\nAuthentication, verify the bidirectional link\nAcknowledgment\nAuthentication\nspoofing\nTransport\nFlooding\nClient puzzles\nDesynchronization\nAuthentication\nSource: Y. Wang, G. Attebury, and B. Ramamurthy, IEEE Communications Surveys and\nTutorials, Vol. 8, no. 2, pp. 2–23, 2006. With permission.\n14.4\nCryptography in WSNs\nSelecting the most appropriate cryptographic method is vital in WSNs as\nall security services are ensured by cryptography. Cryptographic methods\nused in WSNs should meet the constraints of sensor nodes and be evaluated\nby code size, data size, processing time, and power consumption. In this\nsection, we focus on the selection of cryptography in WSNs. We discuss\npublic key cryptography first, followed by symmetric key cryptography.\n14.4.1\nPublic Key Cryptography in WSNs\nMany researchers believe that the code size, data size, processing time, and\npower consumption make it undesirable for public key algorithm tech-\nniques, such as the Diffie–Hellman key agreement protocol [29] or RSA\nsignatures [30], to be employed in WSNs.\nPublic key algorithms such as RSA are computationally intensive and\nusually execute thousands or even millions of multiplication instructions\n" }, { "page_number": 463, "text": "452\n■\nSecurity in Wireless Mesh Networks\nto perform a single security operation. Further, a microprocessor’s public\nkey algorithm efficiency is primarily determined by the number of clock\ncycles required to perform a multiply instruction [31]. Brown et al. found\nthat public key algorithms such as RSA usually require on the order of tens\nof seconds and up to minutes to perform encryption and decryption op-\nerations in constrained wireless devices, which exposes a vulnerability to\nDoS attacks [32]. On the other hand, Carman et al. found that it usually\ntakes a microprocessor thousands of nano-joules to do a simple multiply\nfunction with a 128-bit result [31]. In contrast, symmetric key cryptographic\nalgorithms and hash functions consume much less computational energy\nthan public key algorithms. For example, the encryption of a 1024-bit block\nconsumes approximately 42 mJ on the MC68328 DragonBall processor us-\ning RSA, and the estimated energy consumption for a 128-bit AES block is\na much lower at 0.104 mJ [31].\nRecent studies have shown that it is feasible to apply public key cryp-\ntography to sensor networks by using the right selection of algorithms and\nassociated parameters, optimization, and low power techniques [6,7,33].\nThe investigated public key algorithms include Rabin’s Scheme [34], Ntru-\nEncrypt [35], RSA [30], and Elliptic Curve Cryptography (ECC) [36,37]. Most\nstudies in the literature focus on RSA and ECC algorithms. The attraction\nof ECC is that it offers equal security for a far smaller key size, thereby\nreducing processing and communication overhead. For example, RSA with\n1024-bit keys (RSA-1024) provides a currently accepted level of security for\nmany applications and is equivalent in strength to ECC with 160-bit keys\n(ECC-160) [38]. To protect data beyond the year 2010, RSA Security rec-\nommends RSA-2048 as the new minimum key size, which is equivalent to\nECC with 224-bit keys (ECC-224) [39]. Table 14.3 summarizes the execution\nTable 14.3\nPublic Key Cryptography: Average ECC and RSA\nExecution Times\nAlgorithm\nOperation Time (s)\nECC secp160r1\n0.81\nECC secp224r1\n2.19\nRSA-1024 public key e = 216 + 1\n0.43\nRSA-1024 private key w. CRTa\n10.99\nRSA-2048 public key e = 216 + 1\n1.94\nRSA-2048 private key w. CRT\n83.26\na Chinese Remainder Theory.\nSource: Y. Wang, G. Attebury, and B. Ramamurthy, IEEE Commu-\nnications Surveys and Tutorials, Vol. 8, no. 2, pp. 2–23, 2006.\nWith permission.\n" }, { "page_number": 464, "text": "Security in Wireless Sensor Networks\n■\n453\ntime of ECC and RSA implementations on an Atmel ATmega128 processor\n(used by Mica2 mote) [6]. The execution time is measured on average for\na point multiplication in ECC and a modular exponential operation in RSA.\nECC secp160r1 and secp224r1 are two standardized elliptic curves defined\nin [40]. As shown in Table 14.3, by using the small integer e = 216 + 1 as\nthe public key, RSA public key operation is slightly faster than ECC point\nmultiplication. However, ECC point multiplication outperforms RSA private\nkey operation by an order of magnitude. The RSA private key operation,\nwhich is too slow, limits its use in a sensor node. ECC has no such issues\nbecause both the public key operation and private key operation use the\nsame point multiplication operations.\nWander et al. investigated the energy cost of authentication and key\nexchange based on RSA and ECC cryptography on an Atmel ATmega128\nprocessor [7]. The result is shown in Table 14.4. The ECC-based signature\nis generated and verified with the Elliptic Curve Digital Signature Algorithm\n(ECDSA) [41]. The key exchange protocol is a simplified version of the SSL\nhandshake, which involves two parties: a client initiating the communica-\ntion and a server responding to the initiation [42]. The WSN is assumed\nto be administered by a central point with each sensor having a certificate\nsigned by the central point’s private key using an RSA or ECC signature. In\nthe handshake process, the two parties verify each other’s certificate and\nnegotiate the session key to be used in the communication. As Table 14.4\nshows, compared with RSA cryptography at the same security level, ECDSA\nsignatures are significantly cheaper than RSA signatures and ECDSA verifi-\ncations are within reasonable range of RSA verifications. Further, the ECC-\nbased key exchange protocol outperforms the RSA-based key exchange\nprotocol at the server side, and there is almost no difference in the energy\ncost for these two key exchange protocols at the client side. In addition, the\nTable 14.4\nPublic Key Cryptography: Average\nEnergy Costs of Digital Signature and Key Exchange\nComputations [mJ]\nSignature\nKey Exchange\nAlgorithm\nSign\nVerify\nClient\nServer\nRSA-1024\n304\n11.9\n15.4\n304\nECDSA-160\n22.82\n45.09\n22.3\n22.3\nRSA-2048\n2302.7\n53.7\n57.2\n2302.7\nECDSA-224\n61.54\n121.98\n60.4\n60.4\nSource: Y. Wang, G. Attebury, and B. Ramamurthy,\nIEEE Communications Surveys and Tutorials, Vol. 8,\nno. 2, pp. 2–23, 2006. With permission.\n" }, { "page_number": 465, "text": "454\n■\nSecurity in Wireless Mesh Networks\nrelative performance advantage of ECC over RSA increases as the key size\nincreases in terms of the execution time and energy cost. Table 14.3 and\nTable 14.4 indicate that ECC is more appropriate than RSA for use in sensor\nnetworks.\nThe implementation of RSA and ECC cryptography on Mica2 motes\nfurther proved that a public key-based protocol is viable for WSNs. Two\nmodules, TinyPK [43], based on RSA, and TinyECC [44], based on ECC,\nhave been designed and implemented on Mica2 motes using the TinyOS\ndevelopment environment. Similar work was also conducted by Malan et al.\non ECC cryptography using a Mica2 mote [45]. In their work, ECC was used\nto distribute a single symmetric key for the link layer encryption provided\nby the TinySec module [46].\nWhile public key cryptography may be possible in sensor nodes, the\nprivate key operations are still expensive. The assumptions in [33,45] may\nnot be satisfied in some applications. For example, the work in [33,45]\nconcentrated on the public key operations only, assuming the private key\noperations will be performed by a base station or a third party. By se-\nlecting appropriate parameters, for example, using the small integer e =\n216 + 1 as the public key, the public key operation time can be extremely\nfast while the private key operation time does not change. The limita-\ntion of private key operation occurring only at a base station makes many\nsecurity services using public key algorithms not available under these\nschemes. Such services include peer-to-peer authentication and secure data\naggregation.\nIn contrast, Table 14.5 and Table 14.6 show the execution time and\nenergy cost of two symmetric cryptography protocols on an Atmel AT-\nmega128 processor. In Table 14.5, the execution time was measured on a\n64-bit block using an 80-bit key. From the table we can see that symmetric\nkey cryptography is faster and consumes less energy when compared to\npublic key cryptography. In the remaining section, we focus on symmetric\nkey cryptography.\nTable 14.5\nSymmetric Key Cryptography:\nAverage RC5 and Skipjack Execution Times\nAlgorithm\nOperation Time (ms)\nSkipjack (C) [47]\n0.38\nRC5 (C, assembly) [48]\n0.26\nSource: Y. Wang, G. Attebury, and B. Ramamurthy,\nIEEE Communications Surveys and Tutorials, Vol. 8,\nno. 2, pp. 2–23, 2006. With permission.\n" }, { "page_number": 466, "text": "Security in Wireless Sensor Networks\n■\n455\nTable 14.6\nSymmetric Key Cryptography: Average Energy\nNumbers for AES and SHA-1\nAlgorithm\nEnergy\nSHA-1 (C) [49]\n5.9 μJ/byte\nAES-128 Enc/Dec (assembly) [50]\n1.62/2.49 μJ/byte\nSource: Y. Wang, G. Attebury, and B. Ramamurthy, IEEE Com-\nmunications Surveys and Tutorials, Vol. 8, no. 2, pp. 2–23,\n2006. With permission.\n14.4.2\nSymmetric Key Cryptography in WSNs\nThe constraints on computation and power consumption in sensor nodes\nlimit the application of public key cryptography in WSNs. Thus, most re-\nsearch studies focus on symmetric key cryptography in sensor networks.\nFive popular encryption schemes, RC4 [51], RC5 [48], IDEA [51], SHA-\n1 [49], and MD5 [51,52], were evaluated on six different microprocessors\nranging in word size from 8-bit (Atmel AVR) to 16-bit (Mitsubishi M16C)\nto 32-bit widths (StrongARM, XScale) in [53]. The execution time and code\nmemory size were measured for each algorithm and platform. The experi-\nments indicated uniform cryptographic cost for each encryption class and\neach architecture class. The impact of caches was negligible while Instruc-\ntion Set Architecture (ISA) support is limited to specific effects on certain\nalgorithms. Moreover, hashing algorithms (MD5, SHA-11) incur almost an\norder of magnitude higher overhead than encryption algorithms (RC4, RC5,\nand IDEA).\nIn [54], Law et al. evaluated two symmetric key algorithms: RC5 and\nTEA [55]. They further evaluated six block ciphers including RC5, RC6 [56],\nRijndael [50], MISTY1 [57], KASUMI [58], and Camellia [59] on IAR Systems’\nMSP430F149 in [60]. The benchmark parameters were code, data memory,\nand CPU cycles. The evaluation results showed that Rijndael is suitable for\nhigh security and energy efficiency requirements and MISTY1 is suitable for\ngood storage and energy efficiency. The evaluation results in [60] disagreed\nwith the work in [8] in which RC5 was selected as the encryption/decryption\nscheme, and with the work in [22] in which RC6 was selected. The work\nin [60] provides a good resource for deciding which symmetric algorithm\nshould be adopted in sensor networks.\nThe performance of symmetric key cryptography is mainly decided by\nthe following factors:\n" }, { "page_number": 467, "text": "456\n■\nSecurity in Wireless Mesh Networks\n■\nEmbedded data bus width: Many encryption algorithms prefer 32-bit\nword arithmetic, but most embedded processors usually use an 8-\nor 16-bit wide data bus.\n■\nInstruction set: The ISA has specific effects on certain algorithms. For\nexample, most embedded processors do not support the variable-bit\nrotation instruction like ROL (rotate bits left) of the Intel architecture\nwhich greatly improves the performance of RC5.\nDue to the constraints in sensor nodes, symmetric key cryptography is\npreferred in a WSN.\n14.4.3\nOpen Research Issues\nSelecting the appropriate cryptography method for sensor nodes is funda-\nmental to provide security services in WSNs. However, the decision de-\npends on the computation and communication capability of the sensor\nnodes. Open research issues range from cryptographic algorithms to hard-\nware design as described below:\n■\nRecent studies on public key cryptography have demonstrated\nthat public key operations may be practical in sensor networks.\nHowever, private key operations are still too expensive in terms\nof computation and energy cost to accomplish in a sensor node.\nThe application of private key operations to sensor nodes needs to\nbe studied further.\n■\nSymmetric key cryptography is superior to public key cryptography\nin terms of speed and low energy cost. However, the key distribu-\ntion schemes based on symmetric key cryptography are not perfect.\nEfficient and flexible key distribution schemes need to be designed.\n■\nIt is also likely that more powerful motes will need to be designed\nto support the increasing requirements on computation and com-\nmunication in sensor nodes.\n14.5\nKey Management Protocols\nKey management is a core mechanism to ensure the security of network\nservices and applications in WSNs. The goal of key management is to es-\ntablish required keys between sensor nodes which must exchange data.\nFurther, a key management scheme should also support node addition and\nrevocation while working in undefined deployment environments. Due to\nthe constraints on sensor nodes, key management schemes in WSNs have\nmany differences with the schemes in ad hoc networks.\nAs shown in Section 14.4, public key cryptography suffers from limita-\ntions in WSNs. Thus, most proposed key management schemes are based\n" }, { "page_number": 468, "text": "Security in Wireless Sensor Networks\n■\n457\nProbability of key sharing\nNetwork structure\nKey management protocols in WSNs\nCentralized\nkey scheme\nDistributed\nkey scheme\nProbabilistic\nkey scheme\nLKHW [62],\nLeap [63],\nBrosk [64],\nCDTKeying [65],\nIOS/DMBS [66]\nDeterministic\nkey scheme\nRandom key schemes\n[67, 68, 69, 70, 71, 72,\n73, 74]\nLeap [63], Brosk [64],\nCDTKeying [65],\nIOS/DMBS [66],\nRandom key schemes [67\n68, 69, 70, 71, 72, 73, 74]\nLKHW[62]\nFigure 14.2\nKey management protocols in WSNs: A taxonomy. (From Y. Wang, G.\nAttebury, and B. Ramamurthy, IEEE Communications Surveys and Tutorials, Vol. 8,\nno. 2, pp. 2–23, 2006. With permission.)\non symmetric key cryptography. Further, a straight pairwise private key\nsharing scheme between every pair of nodes is also impractical in WSNs.\nA pairwise private key sharing scheme requires pre-distribution and stor-\nage of n −1 keys in each node, where n is the number of nodes in a\nsensor network. Due to the large amount of memory required, pairwise\nschemes are not viable when the network size is large. Moreover, most key\npairs would be unusable because direct communication is possible only\namong neighboring nodes. This scheme is also not flexible for node addi-\ntion and revocation. In this section, we discuss key management protocols\nin WSNs. Another investigation of key management mechanisms for WSNs\ncan be found in [61].\nFigure 14.2 shows a taxonomy of key management protocols in WSNs.\nAccording to the network structure, the protocols can be divided into cen-\ntralized key schemes and distributed key schemes. According to the prob-\nability of key sharing between a pair of sensor nodes, the protocols can\nbe divided into probabilistic key schemes and deterministic key schemes.\nIn this section, we present a detailed overview of the main key manage-\nment protocols in WSNs. We start with key management protocols based\non network structure.\n14.5.1\nNetwork Structure-Based Key Management Protocols\nThe underlying network structure plays a significant role in the operation of\nkey management protocols. According to the structure, the protocols can\n" }, { "page_number": 469, "text": "458\n■\nSecurity in Wireless Mesh Networks\nbe divided into two categories: centralized key schemes and distributed\nkey schemes.\n14.5.1.1\nCentralized Key Management Schemes\nIn a centralized key scheme, there is only one entity, which is often called\na key distribution center (KDC), controlling the generation, regeneration,\nand distribution of keys. The only proposed centralized key management\nscheme for WSNs in the current literature is the LKHW scheme, which is\nbased on Logical Key Hierarchy (LKH) [62]. In this scheme, the base station\nis treated as a KDC and all keys are logically distributed in a tree rooted at\nthe base station.\nThe central controller does not have to rely on any auxiliary entity to\nperform access control and key distribution. However, with only one man-\naging entity, the central server is a single point of failure. The entire network\nand its security will be affected if there is a problem with the controller.\nDuring the time when the controller is not working, the network becomes\nvulnerable as keys are not generated, regenerated, and distributed. Further-\nmore, the network may become too large to be managed by a single entity,\nthus affecting scalability.\n14.5.1.2\nDistributed Key Management Schemes\nIn the distributed key management approaches, different controllers are\nused to manage key generation, regeneration, and distribution, minimizing\nthe risk of failure and allowing for better scalability. In this approach, more\nentities are allowed to fail before the whole network is affected.\nMost proposed key management schemes are distributed schemes. These\nschemes also fall into deterministic and probabilistic categories, which are\ndiscussed in detail in the following sub-section.\n14.5.2\nKey Management Protocols Based on the Probability\nof Key Sharing\nIn the remainder of this section, we present the key management protocols\nbased on the probability of key sharing between a pair of sensor nodes.\nWe first discuss deterministic approaches and then discuss probabilistic\napproaches.\n14.5.2.1\nDeterministic Approaches\nZhu et al. proposed a key management protocol, Localized Encryption and\nAuthentication Protocol (LEAP), for sensor networks in [63]. LEAP supports\nthe establishment of four types of keys for each sensor node:\n" }, { "page_number": 470, "text": "Security in Wireless Sensor Networks\n■\n459\n■\nAn individual key shared with the base station (pre-distributed)\n■\nA group key shared by all the nodes in the network (pre-distributed)\n■\nPairwise keys shared with immediate neighboring nodes\n■\nA cluster key shared with multiple neighboring nodes\nThe pairwise keys shared with immediate neighboring nodes are used to\nprotect peer-to-peer communication and the cluster key is used for local\nbroadcast. The pairwise keys can be set up as follows: in the key pre-\ndistribution stage, each sensor node is loaded with an initial key K I and\neach node A generates a master key K A = fK I (A), where f is a pseudo-\nrandom function. Then, in the neighbor discovery stage, A broadcasts a\nHello message and expects an acknowledgment from neighboring nodes,\ne.g., node B:\nA −→∗: A\nB −→A : B, M AC(K B, A|B)\nNode A computes its pairwise key with B, K AB = fK B(A) and node B\nknows A, K B and can also compute K AB in the same way. Then, K AB\nserves as their pairwise key.\nCluster key establishment follows the pairwise key establishment phase.\nSuppose node A wants to establish a cluster key with all its immediate\nneighbors B1, B2, . . . , Bm. Node A first generates a random key K c\nA, then\nencrypts this key with the pairwise key shared with each neighbor, and\nfinally transmits the encrypted key to each neighbor Bi, where 1 ≤i ≤m.\nA −→Bi :\n\u0002K c\nA\n\u0003\nK ABi\nLEAP uses unicast for key exchange. Notice that most of the proposed\nsecurity protocols were based on point-to-point handshaking procedures\nto negotiate session keys. Lai et al. proposed a BROadcast Session Key\n(BROSK) negotiation protocol [64]. BROSK assumes a master key is shared\nby all nodes in the network. To establish a session key K with its neighbors,\nsuch as node B, a sensor node A broadcasts a key negotiation message:\nA −→∗: IDA|NA, M AC(K, IDA|NA)\nB −→∗: IDB|NB, M AC(K, IDB|NB)\nA and B will receive the broadcast message. They can verify the message\nusing the master key K and both A and B can calculate the shared session\nkey:\nK AB = M AC(K, NA|NB)\n" }, { "page_number": 471, "text": "460\n■\nSecurity in Wireless Mesh Networks\nBROSK therefore establishes pairwise session keys between every two\nneighboring nodes. It is both scalable and energy efficient.\nCamtepe and Yener proposed a deterministic key distribution scheme\nfor WSNs using Combinatorial Design Theory [65]. The Combinatorial De-\nsign Theory based pairwise key pre-distribution (CDTKeying) scheme is\nbased on block design techniques in combinatorial design theory. It em-\nploys symmetric and generalized quadrangle design techniques. The\nscheme uses a finite projective plane of order n (for prime power n) to\ngenerate a symmetric design with parameters n2 + n + 1, n + 1, 1. The de-\nsign supports n2 + n + 1 nodes and uses a key pool of size n2 + n + 1. It\ngenerates n2 + n + 1 key chains of size n + 1 where every pair of key\nchains has exactly one key in common, and every key appears in exactly\nn + 1 key-chains. After the deployment, every pair of nodes finds exactly\none common key. Thus, the probability of key sharing among a pair of\nsensor nodes is 1. The disadvantage of this solution is that the parameter\nn has to be a prime power, thus indicating that not all network sizes can\nbe supported for a fixed key chain size.\nLee and Stinson proposed two combinatorial design theory based de-\nterministic schemes: ID-based one-way function scheme (IOS) and deter-\nministic multiple space Blom’s scheme (DMBS) [66]. They further discussed\nthe use of combinatorial set systems in the design of deterministic key\npre-distribution schemes for WSNs in [67].\n14.5.2.2\nProbabilistic Approaches\nMost proposed key management schemes in WSNs are probabilistic and\ndistributed schemes.\nEschenauer and Gligor introduced a key pre-distribution scheme for\nsensor networks which relies on probabilistic key sharing among the nodes\nof a random graph [68]. The scheme consists of three phases: key pre-\ndistribution, shared key discovery, and path key establishment. In the key\npre-distribution phase, each sensor is equipped with a key ring held in the\nmemory. The key ring consists of k keys which are randomly drawn from\na large pool of P keys. The association information of the key identifiers in\nthe key ring and sensor identifier is also stored at the base station. Further,\nthe authors assumed that each sensor shares a pairwise key with the base\nstation. In the shared key discovery phase, each sensor discovers its neigh-\nbors within wireless communication range with which it shares keys. Two\nmethods to accomplish this are suggested in [68]. The simplest method is\nfor each node to broadcast a list of identifiers of the keys in their key ring in\nplaintext allowing neighboring nodes to check whether they share a key.\nHowever, the adversary may observe the key-sharing patterns among sen-\nsors in this way. The second method uses the challenge–response technique\nto hide key-sharing patterns among nodes from an adversary. For every Ki\n" }, { "page_number": 472, "text": "Security in Wireless Sensor Networks\n■\n461\non a key ring, each node could broadcast a list α, E Ki(α), i = 1, . . . , k\nwhere α is a challenge. The decryption of E Ki(α) with the proper key by\na recipient would reveal the challenge and establish a shared key with the\nbroadcasting node. This method requires the challenge α be well known in\nthe sensor network, allowing the recipient with the proper key to discover\nthe challenge.\nFinally, in the path key establishment phase, a path key is assigned for\nthose sensor nodes within wireless communication range and not sharing\na key, but connected by two or more links at the end of the second phase.\nIf a node is compromised, the base station can send a message to all other\nsensors to revoke the compromised node’s key ring. Re-keying follows\nthe same procedure as revocation. The messages from the base station are\nsigned by the pairwise key shared by the base station and sensor nodes,\nthus ensuring that no adversary can forge a base station. If a node is com-\npromised, the attacker has a probability of approximately k/P to attack any\nlink successfully. Because k ≪P, it only affects a small number of sensor\nnodes.\nInspired by the work in [68], which we call the basic random key scheme\nin the following section, additional random key pre-distribution schemes\nhave been proposed in [69–74].\nIn the basic random key scheme, any two neighboring nodes need to\nfind a single common key from their key rings to establish a secure link\nin the key setup phase. However, Chan et al. observed that increasing\nthe amount of key overlap in the key ring can increase the resilience of\nthe network against node capture [69]. Thus, they proposed a q-composite\nkeying scheme. It is required to share at least q common keys in the key\nsetup phase to build a secure link between any two neighboring nodes.\nFurther, they introduced a key update phase to enhance the basic random\nkey scheme. Suppose A has a secure link to B after the key setup phase\nand the secure key is k from the key pool P. Because k may be residing\nin the key ring memory of some other nodes in the network, the security\nof the link between A and B is jeopardized if any of those nodes are\ncaptured. Thus, it is better to update the communication key between A\nand B instead of using a key in the key pool. To address the problem,\nthey presented a multipath key reinforcement for the key update. Assume\nthere are j disjoint paths between A and B. A generates j random values\nv1, v2, . . . , v j and then routes each random value along a different path to\nB. When B has received all j keys, the new link key can be computed by\nboth A and B as:\nk′ = k ⊕v1 ⊕v2 ⊕. . . ⊕v j\nThe adversary has to eavesdrop on all j paths if he wants to reconstruct the\ncommunication key. This security enhancement comes at the cost of more\n" }, { "page_number": 473, "text": "462\n■\nSecurity in Wireless Mesh Networks\ncommunication overhead needed to find multiple disjoint paths. Further,\nChan et al. [69] also developed a random-pairwise key scheme for node-\nto-node authentication.\nBlundo et al. presented a polynomial-based key pre-distribution pro-\ntocol for group key pre-distribution in [75], which can also be adapted\nto sensor networks. The key setup server randomly generates a bivariate\nt-degree polynomial f (x, y) = \u0004t\ni, j=0 aij xi y j over a finite field Fq where\nq is a prime number that is large enough to accommodate a cryptographic\nkey such that it has the property of f (x, y) = f (y, x). For each sensor i,\nthe setup server computes a polynomial share of f (x, y), that is, f (i, y).\nFor any two sensor nodes i and j, node i can compute the common key\nf (i, j) by evaluating f (i, y) at point j, and node j can compute the same\nkey f ( j, i) = f (i, j) by evaluating f ( j, y) at point i. In this approach,\neach sensor node i needs to store a t-degree polynomial f (i, x), which\noccupies (t + 1) log q storage space. This scheme is unconditionally secure\nand t-collusion resistant. However, the storage cost for a polynomial share\nis exponential in terms of the group size, making it prohibitive in sensor\nnetworks.\nInspired by the work of [68,69,75], Liu and Ning proposed a polyno-\nmial pool-based key pre-distribution scheme in [70], which also includes\nthree phases: setup, direct key establishment, and path key establishment.\nIn the setup phase, the setup server randomly generates a set F of bivari-\nate t-degree polynomials over the finite field Fq. For each sensor node,\nthe setup server picks a subset of polynomials Fi ⊆F and assigns the\npolynomial shares of these polynomials to node i. In the direct key es-\ntablishment stage, the sensor nodes find a shared polynomial with other\nsensor nodes and then establish a pairwise key using the polynomial-based\nkey pre-distribution scheme discussed in [75]. The path key establishment\nphase is similar to that in the basic random key scheme. Further, the pro-\nposed framework allows for the study of multiple instantiations of possible\npairwise key establishment schemes. Two of the possible instantiations, the\nkey pre-distribution scheme based on random subset assignment and the\ngrid-based key pre-distribution scheme, are also presented and analyzed\nin the paper.\nSimilar to [70], Du et al. presented another pairwise key pre-distribution\nscheme in [72] which uses Blom’s method [76]. The key difference between\n[70] and [72] is that the scheme in [70] is based on a set of bivariate t-degree\npolynomials and Du’s scheme is based on Blom’s method. The proposed\nscheme allows any pair of nodes in a network to be able to find a pairwise\nsecret key. As long as no more than λ nodes are compromised, the network\nis perfectly secure (which is called the λ-secure property). To use Blom’s\nmethod, during the pre-deployment phase, the base station first constructs\na (λ + 1) × N matrix G over a finite field G F (q), where N is the size of\nthe network and G is considered to be public information. Then the base\n" }, { "page_number": 474, "text": "Security in Wireless Sensor Networks\n■\n463\nstation creates a random (λ + 1) × (λ + 1) symmetric matrix D over G F (q),\nand computes an N × (λ + 1) matrix A = (D · G)T , where (D · G)T is the\ntranspose of D · G. Matrix D needs to be kept secret, and should not be\ndisclosed to adversaries. It is easy to verify that A· G is a symmetric matrix.\nA · G = (D · G)T · G = GT · DT · G = GT · D · G\n= (A · G)T\nThus, we know Kij = K ji. The idea is to use Kij (or K ji) as the pairwise\nkey between node i and node j. To carry out the above computation, in\nthe pre-distribution phase, for any sensor k = 1, . . . , N:\n■\nStore the k th row of matrix A at node k.\n■\nStore the k th column of matrix G at node k.\nTherefore, when nodes i and j need to find the pairwise key between\nthem, they first exchange their columns of G, and then compute Kij and\nKji, respectively, using their private rows of A.\nIn the proposed scheme in [72], each sensor node is loaded with G\nand τ distinct D matrices drawn from a large pool of ω symmetric matrices\nD1, . . . , Dω of size (λ + 1) × (λ + 1). For each Di, calculate the matrix\nAi = (Di · G)T and store the jth row of Ai at this node. After deployment,\neach node needs to discover whether it shares any space with neighbors.\nIf they find out that they have a common space, the nodes can follow\nBlom’s method to build a pairwise key. The scheme in [72] is scalable and\nflexible. Moreover, it is substantially more resilient against node capture as\ncompared to [70].\nHwang et al. extended the basic random key scheme and proposed\na cluster key grouping scheme [74]. They further analyzed the trade-offs\ninvolved between energy, memory, and security robustness.\nNotice that location information helps to avoid unnecessary key as-\nsignments and thus improve the performance of sensor networks, such as\nconnectivity, memory usage, and network resilience against node capture.\nTaking this into account, two random key pre-distribution schemes were\nproposed in [73] and [77]. Although the presented schemes show improved\nperformance, the deployment information, such as location, is required\nwhen sensors are deployed.\nThe above-mentioned schemes are classified and compared in Table 14.7.\n14.5.3\nOpen Research Issues\nAlthough some key management protocols have been proposed for sensor\nnetworks, the design of key management protocols is still largely open to\nresearch. Open research issues include the following:\n" }, { "page_number": 475, "text": "464\n■\nSecurity in Wireless Mesh Networks\nTable 14.7\nClassification and Comparison of Key Management Protocols in WSNs\nMaster Pairwise Path\nCluster\nProcessing\nComm.\nStorage\nProtocol\nRef.\nKey\nKey\nKey\nKey\nScalability Resiliency\nLoad\nLoad\nLoad\nI\nAll pairwise\n—\nn/a\nYes\nNo\nNo\nLow\nLow\nLow\nLow\nHigh\nLEAP\n[63]\nYes\nYes\nYes\nYes\nGood\nLow\nLow\nLow\nLow\nBROSK\n[64]\nYes\nYes\nNo\nNo\nGood\nLow\nLow\nLow\nLow\nLKHW\n[62]\nYes\nYes\nNo\nYes\nLimited\nLow\nLow\nLow\nLow\nCDTKeying\n[65]\nn/a\nYes\nNo\nNo\nGood\nGood\nMedium\nMedium\nHigh\nIOS & DMBS\n[66]\nn/a\nYes\nNo\nNo\nGood\nGood\nMedium\nMedium\nHigh\nII\n[68]\nn/a\nYes\nYes\nNo\nGood\nGood\nMedium\nMedium\nHigh\nq-composite\n[69]\nn/a\nYes\nYes\nNo\nGood\nGood\nMedium\nMedium\nHigh\nPolynomial based\n[70]\nn/a\nYes\nYes\nNo\nGood\nGood\nMedium\nMedium\nHigh\nBlom based\n[72]\nn/a\nYes\nYes\nNo\nGood\nGood\nMedium\nMedium\nHigh\nDeployment knowledge based\n[73]\nn/a\nYes\nYes\nNo\nGood\nGood\nMedium\nMedium\nMedium\nCluster key grouping\n[74]\nn/a\nYes\nYes\nNo\nGood\nGood\nMedium\nMedium\nHigh\nLocation based\n[77]\nn/a\nYes\nYes\nNo\nGood\nGood\nMedium\nMedium\nMedium\nNote: Category I denotes deterministic approaches and category II denotes probabilistic approaches. Master key is the key shared by all the\nnodes in the network. Pairwise key is the key shared between two neighboring nodes. Path key denotes the key shared between any two nodes\nwhich need exchange data, but do not share a pairwise key. Cluster key denotes the common key shared by all cluster members.\n" }, { "page_number": 476, "text": "Security in Wireless Sensor Networks\n■\n465\n■\nThe proposed key management protocols discussed in this section\nemploy different strategies on the trade-off between memory, pro-\ncessing, and communication overhead. These protocols could be\nimproved and new key management protocols need to be designed.\n■\nAll key management protocols discussed in the literature so far are\nbased on symmetric key cryptography. Recent progress in public\nkey cryptography has shown that public key cryptography may be\nsuitable for sensor networks. Key management schemes based on\npublic key cryptography need to be designed.\n■\nCurrent proposed key management schemes assume that the base\nstation is trustworthy. However, there may be situations, such as in\nthe battlefield, where the security of a base station needs to be con-\nsidered. New schemes need to be designed considering the security\nof base stations.\n14.6\nSecure Routing Protocols\nMany routing protocols have been specifically designed for WSNs. These\nrouting protocols can be divided into three categories according to the net-\nwork structure: flat-based routing, hierarchical-based routing, and location-\nbased routing [78]. In flat-based routing, all nodes are typically assigned\nequal roles or functionality. In hierarchical-based routing, nodes play dif-\nferent roles in the network. In location-based routing, sensor node positions\nare used to route data in the network. Although many sensor network rout-\ning protocols have been proposed in the literature, few of them have been\ndesigned with security as a goal. Lacking security services in the routing\nprotocols, WSNs are vulnerable to many kinds of attacks.\nMost network layer attacks against sensor networks fall into one of the\ncategories described in Section 14.3.3, namely:\n■\nSpoofed, altered, or replayed routing information\n■\nSelective forwarding\n■\nSinkhole\n■\nSybil\n■\nWormholes\n■\nHello flood attacks\n■\nAcknowledgment spoofing\nThese attacks may be applied to compromise the routing protocols in a\nsensor network. For example, directed diffusion is a flat-based routing\nalgorithm for drawing information from a sensor network [79]. In directed\ndiffusion, sensors measure events and create gradients of information in\ntheir respective neighboring nodes. The base station requests data by\n" }, { "page_number": 477, "text": "466\n■\nSecurity in Wireless Mesh Networks\nbroadcasting interest which describes a task to be conducted by the net-\nwork. The interest is diffused through the network hop by hop, and broad-\ncasted by each node to its neighbors. As the interest is propagated through-\nout the network, gradients are set up to draw data satisfying the query to-\nward the requesting node. Each sensor that receives the interest sets up a\ngradient toward the sensor nodes from which it received the interest. This\nprocess continues until gradients are set up from the sources back to the\nbase station. Interests initially specify a low rate of data flow, but once a\nbase station starts receiving events, it will reinforce one or more neighbor-\ning nodes to request higher data rate events. This process proceeds recur-\nsively until it reaches the nodes generating events, causing them to generate\nevents at a higher data rate. Paths may also be negatively reinforced. Di-\nrected diffusion is vulnerable to many kinds of attacks if authentication is\nnot included in the protocol [26]. For example, it is easy for an adversary\nto add himself onto the path taken by a flow of events, as described in the\nfollowing:\n■\nThe adversary can influence the path by spoofing positive reinforce-\nments. After receiving and rebroadcasting an interest, an adversary\ncould strongly reinforce the nodes to which the interest was sent\nwhile spoofing high rate, low latency events to the nodes from which\nthe interest was received.\n■\nThe adversary can replay the interests intercepted from a legitimate\nbase station and list himself as a base station. All events satisfying\nthe interest will then be sent to both the adversary and the legitimate\nbase station.\nBy using the attacks above, the adversary can add himself onto the path\nand thus gain full control of the flow. The adversary can eavesdrop, mod-\nify, and selectively forward packets of his choosing. He can drop all for-\nwarded packets and act as a sinkhole. Further, a laptop-class adversary\ncan exert great influence on the topology by using a wormhole attack.\nThe adversary creates a tunnel between a node located near a base station\nand a node located close to where events are likely to be generated. By\nspoofing positive or negative reinforcements, the adversary can push data\nflows away from the base station and toward the nodes selected by the\nadversary.\nHierarchical and location-based routing protocols not incorporating\nsecurity services are also vulnerable to many attacks [26]. For example,\nlocation-based routing protocols such as Geographic and Energy Aware\nRouting (GEAR) [80] require location information to be exchanged between\nneighbors. However, location information can be misrepresented. Regard-\nless of the adversary’s actual location, he may advertise false position data\nto place himself on the path of a known flow. Once on that path, the\n" }, { "page_number": 478, "text": "Security in Wireless Sensor Networks\n■\n467\nadversary can mount selective forwarding and Sybil attacks in the data\nflows. Simulations in [81] found that such attacks have great influence on\nthe overall ratio of successfully delivered messages in the network.\nSecure routing in ad hoc networks is similar to that in sensor networks\nand has been well studied in literature [14]. However, the defense mecha-\nnisms developed for ad hoc networks cannot be directly applied to sensor\nnetworks because of the differences between sensor and ad hoc networks\ndiscussed in Section 14.1.\nIdeally, a secure routing protocol should guarantee the integrity, authen-\ntication, and availability of messages in the presence of adversaries of arbi-\ntrary power. In the presence of only outsider adversaries, it is conceivable\nto achieve these idealized goals. However, in the presence of compromised\nnodes or insider adversaries, especially those with laptop-class capabilities,\nit is most likely that some if not all of these goals are not fully attainable. In\nthis situation, the best we can hope for is graceful degradation instead of a\ncomplete compromise of the network. To achieve the above goal requires\nthat a routing protocol degrades no faster than a rate approximately propor-\ntional to the ratio of compromised nodes to total nodes in the network [26].\nA secure routing protocol depends on an appropriate key management\nscheme in a WSN, which has been discussed in Section 14.5. Before a\nrouting protocol starts, sensor nodes should have been loaded with proper\nkeys, e.g., the key for confidentiality, authentication, etc. One of the fun-\ndamental security services in sensor networks is broadcast authentication,\nwhich enables the base station to broadcast authenticated data to the entire\nsensor network. In this section, we first discuss the broadcast authentication\nproblem and then review several secure routing schemes.\n14.6.1\nBroadcast Authentication\nPrevious proposals for authenticated broadcast are impractical in WSNs for\nthe following reasons:\n■\nMost proposals rely on public key cryptography for the authentica-\ntion. However, public key cryptography is impractical for WSNs.\n■\nEven one-time signature schemes that are based on symmetric key\ncryptography have too much overhead.\nμTESLA (the “micro” version of the Timed, Efficient, Streaming, Loss-tolerant\nAuthentication protocol) [10] and its extensions [82,83] have been proposed\nto provide broadcast authentication for sensor networks.\nμTESLA is an authenticated broadcast protocol which was proposed\nby Perrig et al. for the SPINS protocol [8]. μTESLA introduces asymmetry\nthrough a delayed disclosure of symmetric keys resulting in an efficient\n" }, { "page_number": 479, "text": "468\n■\nSecurity in Wireless Mesh Networks\nbroadcast authentication scheme. μTESLA requires that the base station\nand nodes be loosely time synchronized, and that each node knows an\nupper bound on the maximum synchronization error.\nTo send an authenticated packet, the base station simply computes a\nMAC on the packet with a key that is secret at that point in time. When\na node gets a packet, it can verify that the corresponding MAC key was\nnot yet disclosed by the base station. Because a receiving node is assured\nthat the MAC key is known only by the base station, the receiving node is\nassured that no adversary could have altered the packet in transit. The node\nstores the packet in a buffer. At the time of key disclosure, the base station\nbroadcasts the verification key to all receivers. When a node receives the\ndisclosed key, it can easily verify the correctness of the key. If the key is\ncorrect, the node can now use it to authenticate the packet stored in its\nbuffer.\nEach MAC key is a key from the key chain, generated by a public one-\nway function F . To generate the one-way key chain, the sender chooses\nthe last key Kn from the chain, and repeatedly applies F to compute all\nother keys: Ki = F (Ki+1).\nFigure 14.3 shows an example of μTESLA. The receiver node is loosely\ntime synchronized and knows K0 in an authenticated way. Packets P1 and\nP2 sent in interval 1 contain a MAC with key K1. Packet P3 has a MAC\nusing key K2. If P4, P5, and P6 are all lost, as well as the packet that\ndisclosed key K1, the receiver cannot authenticate P1, P2, and P3. In interval\n4 the base station broadcasts key K2, which the nodes authenticate by\nverifying K0 = F (F (K2)), and hence know also K1 = F (K2), so they can\nauthenticate packets P1, P2 with K1, and P3 with K2.\nSPINS limits the broadcasting capability to only the base station. If a\nnode wants to broadcast authenticated data, the node has to broadcast the\ndata through the base station. The data is first sent to the base station in an\nauthenticated way. It is then broadcasted by the base station.\nTo bootstrap a new receiver, μTESLA depends on a point-to-point au-\nthentication mechanism in which a receiver sends a request message to\nthe base station and the base station replies with a message containing all\nthe necessary parameters. Notice that μTESLA requires the base station to\nF\nF\nF\nF\nP1\nP2\nP3\nP4\nP5\nP6\nP7\nK0\nK1\nK2\nK3\nK4\nTime\nFigure 14.3\nUsing a time-released key chain for source authentication. (From Y.\nWang, G. Attebury, and B. Ramamurthy, IEEE Communications Surveys and Tutorials,\nVol. 8, no. 2, pp. 2–23, 2006. With permission.)\n" }, { "page_number": 480, "text": "Security in Wireless Sensor Networks\n■\n469\nunicast initial parameters to individual sensor nodes, and thus incurs a long\ndelay to boot up a large scale sensor network. Liu and Ning proposed a\nmulti-level key chain scheme for broadcast authentication to overcome this\ndeficiency in [82,83].\nThe basic idea in [82,83] is to predetermine and broadcast the initial\nparameters required by μTESLA instead of using unicast-based message\ntransmission. The simplest way is to pre-distribute the μTESLA parameters\nwith a master key during the initialization of the sensor nodes. As a result,\nall sensor nodes have the key chain commitments and other necessary\nparameters once they are initialized, and are ready to use μTESLA as long\nas the starting time has passed. Furthermore, Liu and Ning introduced a\nmulti-level key chain scheme, in which the higher key chains are used to\nauthenticate the commitments of lower-level ones. However, the multi-level\nkey chain scheme suffers from possible DoS attacks during the commitment\ndistribution stage. Further, none of the μTESLA or multi-level key chain\nschemes is scalable in terms of the number of senders. In [84], a practical\nbroadcast authentication protocol was proposed to support a potentially\nlarge number of broadcast senders using μTESLA as a building block.\nμTESLA provides broadcast authentication for base stations, but is not\nsuitable for local broadcast authentication. This is because μTESLA does not\nprovide immediate authentication. For every received packet, a node has\nto wait for one μTESLA interval to receive the MAC key used in computing\nthe MAC for the packet. As a result, if μTESLA is used for local broadcast\nauthentication, a message traversing l hops will take at least l μTESLA in-\ntervals to arrive at the destination. In addition, a sensor node has to buffer\nall the unverified packets. Both the latency and the storage requirements\nlimit the scheme for authenticating infrequent messages broadcast by the\nbase station. Zhu et al. proposed a one-way key chain scheme for one-hop\nbroadcast authentication in LEAP [63]. In this scheme, every node generates\na one-way key chain of certain length and then transmits the commitment\n(i.e., the first key) of the key chain to each neighbor, encrypted with their\npairwise shared key. Whenever a node has a message to send, it attaches\nto the message to the next authenticated key in the key chain. The authen-\nticated keys are disclosed in reverse order to their generation. A receiving\nneighbor can verify the message based on the commitment or an authen-\nticated key it received from the sending node more recently.\n14.6.2\nSecure Routing\nThe goal of a secure routing protocol is to ensure the integrity, authentica-\ntion, and availability of messages. The proposed secure routing protocols\nfor WSNs in the literature were all based on symmetric key cryptography\nexcept the work in [85], which was based on public key cryptography.\n" }, { "page_number": 481, "text": "470\n■\nSecurity in Wireless Mesh Networks\nSPINS is a suite of security protocols optimized for sensor networks\n[8]. SPINS includes two building blocks: SNEP (Secure Network Encryption\nProtocol) and μTESLA. SNEP provides data confidentiality, two-party data\nauthentication, and data freshness for peer-to-peer communication (node\nto base station). μTESLA provides authenticated broadcast as discussed\nbefore. We discuss SNEP in this sub-section.\nSPINS assumes that each node is pre-distributed with a master key K\nwhich is shared with the base station at creation time. All other keys, in-\ncluding a key Kencr for encryption, a key Kmac for MAC generation, and a\nkey Krand for random number generation, are derived from the master key\nusing a strong one-way function. SPINS uses RC5 for confidentiality. If A\nwants to send a message to base station B, the complete message that A\nsends to B is\nA →B : D⟨KencrC⟩, M AC(Kmac, C|D)⟨KencrC⟩\nwhere D is the transmitted data and C is a shared counter between the\nsender and the receiver for the block cipher in counter mode. The counter\nC is incremented after each message is sent and received in both the sender\nand the receiver side. SNEP also provides a counter exchange protocol to\nsynchronize the counter value in both sides.\nSNEP offers the following properties: semantic security, data authentica-\ntion, replay protection, weak freshness, and low communication overhead.\nSPINS identifies two types of freshness: weak freshness and strong fresh-\nness. Weak freshness provides partial message ordering and carries no delay\ninformation; strong freshness provides a total order on a request-response\npair and allows for delay estimation.\n■\nSemantic security: The counter value is incremented after each mes-\nsage and thus the same message is encrypted differently each time.\n■\nData authentication: A receiver can be assured that the message\noriginated from the claimed sender if the MAC verifies correctly.\n■\nReplay protection: The counter value in the MAC prevents replaying\nold messages.\n■\nWeak freshness: The counter also maintains a message ordering in\nthe receiver side and yields weak freshness. SNEP provides weak\ndata freshness only because there is no absolute assurance to node\nA that a message was created by node B in response to an event in\nnode A.\n■\nLow communication overhead: The counter state is kept at each\nendpoint and does not need to be sent in each message.\nDirected diffusion routing protocol was proposed by Intanagonwiwat\net al. without considering security issues [79]. Pietro et al. proposed an\n" }, { "page_number": 482, "text": "Security in Wireless Sensor Networks\n■\n471\nextension of directed diffusion protocol which provides secure multicasting\nin [62]. The extended scheme, Logical Key Hierarchy for WSNs (LKHW),\nprovides robustness in routing and security and supports both backward\nand forward secrecy for sensor join and leave operations. However, it does\nnot provide data authentication.\nInspired by the work on public key cryptography [6,7,33,43], Du et al.\ninvestigated the public key authentication problem [85]. The use of public\nkey cryptography eases many problems in secure routing, for example, au-\nthentication and integrity. However, before a node A uses the public key\nfrom another node B, A must verify that the public key is actually B’s, i.e.,\nA must authenticate B’s public key; otherwise, man-in-the-middle attacks\nare possible. In general networks, public key authentication involves a sig-\nnature verification on a certificate signed by a trusted third party Certificate\nAuthority (CA) [86]. However, the signature verification operations are still\ntoo expensive for sensor nodes, as depicted in Table 14.3 and Table 14.4.\nDu et al. proposed an efficient alternative that uses only a one-way hash\nfunction for the public key authentication. The proposed scheme can be\ndivided into two stages. In the pre-distribution stage, a Merkle tree R is\nconstructed with each leaf L i corresponding to a sensor node (more infor-\nmation on Merkle trees is given in Section 14.7). Let pki represent node\ni’s public key, V be an internal tree node, and Vlef t and Vright be V’s two\nchildren. The value of an internal tree node is denoted by φ. The Merkle\ntree can then be constructed as follows:\nφ(L i) = h(idi, pki),\nf or i = 1, . . . , N\nφ(V) = h(φ(Vleft) ∥φ(Vright))\nwhere “∥” represents the concatenation of two strings and h is a one-\nway hash function such as MD5 or SHA-1. Let R be the root of the tree.\nEach sensor node v needs to store the root value φ(R) and the sibling\nnode values λ1, . . . , λH along the path from v to R. If node A wants to\nauthenticate B’s public key, B sends its public key pk along with the value\nof λ1, . . . , λH to node A. Then, A can use the same procedure to reconstruct\nthe Merkle tree R′ and calculate the root value φ(R′). A will trust B to\nbe authentic if φ(R′) = φ(R). A sensor node only needs H + 1 storage\nunits for the extra hash values. Based on this scheme, Du et al. further\nextended the idea to reduce the height of the Merkle tree to improve the\ncommunication overhead of the scheme. The proposed scheme is more\nefficient than signature verification on certificates. However, the scheme\nrequires that some hash values be distributed in a pre-distribution stage.\nThis results in some scalability issues when new sensors are added to an\nexisting WSN.\nThe discussion above is summarized in Table 14.8.\n" }, { "page_number": 483, "text": "472\n■\nSecurity in Wireless Mesh Networks\nTable 14.8\nComparison of Secure Routing Protocols\nP2P\nBroadcast\nRef.\nRouting\nConfidentiality\nAuthentication\nAuthentication\nIntegrity\nScalability\nSNEP\n[8]\nFlat\nYes\nYes\nNo\nYes\nGood\nLKHW\n[62]\nFlat\nYes\nNo\nNo\nNo\nLimited\nμTESLA\n[8]\nFlat/hierarchy\nNo\nNo\nYes\nYes\nMedium\nMulti-level key chains\n[82]\nFlat/hierarchy\nNo\nNo\nYes\nYes\nGood\nLEAP\n[63]\nHierarchy\nYes\nYes\nYes\nYes\nMedium\n" }, { "page_number": 484, "text": "Security in Wireless Sensor Networks\n■\n473\n14.6.3\nOpen Research Issues\nThe development of secure routing protocols is challenging because sensor\nnodes are prone to failures, and the topology of a sensor network changes\nfrequently due to node failures and possible mobility. Key open research\nissues include the following:\n■\nThe proposed secure routing protocols for WSNs focus on static\nsensor networks only, ignoring mobility. Secure routing protocols\nfor mobile sensor networks need to be investigated.\n■\nCurrent broadcast authentication schemes such as μTESLA and its\nextensions require the sensor network to be loosely time synchro-\nnized. This requirement is often hard to meet and new techniques\nthat do not require time synchronization are desirable.\n■\nNew schemes with higher scalability and efficiency need to be devel-\noped for the authenticated broadcast protocols. The recent progress\non public key cryptography may facilitate the design of authenti-\ncated broadcast protocols.\n■\nQuality of service in WSNs needs to be evaluated with the addition\nof secure routing services.\n14.7\nSecure Data Aggregation\nData communication constitutes an important share of the total energy con-\nsumption of the sensor network. The simulation in [8] shows that data trans-\nmission accounts for 71 percent of the energy cost of computation and\ncommunication for the SNEP protocol. Thus, data aggregation can greatly\nhelp conserve the scarce energy resources by eliminating redundant data.\nData aggregation (fusion) protocols aim at eliminating redundant data\ntransmitted across the network and are essential for energy-constrained\nWSNs. Traditional data aggregation techniques include simple types of\nqueries such as SUM, COUNT, AVERAGE, and MIN/MAX. Some researchers\nalso extend data aggregation to median, the most frequent (consensus) data\nvalues, a histogram of the data distribution, and range queries [87]. Data\naggregation can be divided into two stages: detection and data fusion.\nIn a WSN, there are usually certain nodes, called aggregators, help-\ning aggregate information requested by queries. When an aggregator node\nis compromised, it is easy for the adversary to inject false data into sen-\nsor networks. Thus, the aggregators are vulnerable to be attacked. An-\nother possible attack is to compromise a sensor node and inject forged\ndata through it. Without authentication, the attackers can fool the aggrega-\ntors into reporting false data to the base station. Secure data aggregation\nrequires authentication, confidentiality, and integrity. Moreover, secure data\n" }, { "page_number": 485, "text": "474\n■\nSecurity in Wireless Mesh Networks\nCiphertext-based\naggregation\nProtocol operations\nPlaintext-based\naggregation\nCDA (93), HSC (94)\nSA (9), SIA (10), SINP (88),\nESPDA (89, 90)\nSDDA (91), WDA (92)\nFigure 14.4\nSecure data aggregation in WSNs: A taxonomy. (From Y. Wang, G. Atte-\nbury, and B. Ramamurthy, IEEE Communications Surveys and Tutorials, Vol. 8, no. 2,\npp. 2–23, 2006. With permission.)\naggregation also requires the cooperation of sensor nodes to identify the\ncompromised sensors.\nHowever, requirements for confidentiality and data aggregation are at\nodds with each other. Confidentiality requires the data to be transmitted\nin ciphertext, data aggregation is usually based on plaintext. A straight-\nforward method is to invoke end-to-end encryption before evoking data\naggregation. However, the trade-off is that the end-to-end encryption and\ndecryption operations consume more energy, which is of great concern\nin WSNs. An alternative way is to provide data aggregation on concealed\ndata, which requires a particular class of encryption transformation. How-\never, this method usually lowers the security level.\nFigure 14.4 shows a taxonomy of secure data aggregation protocols\nin WSNs. According to the protocol operation, secure data aggregation can\nbe classified into two categories: plaintext-based and ciphertext-based. This\nsection reviews the techniques for secure data aggregation.\n14.7.1\nPlaintext-Based Secure Data Aggregation\nHu and Evans proposed a secure aggregation (SA) protocol for WSNs that\nis resilient to both intruder devices and single device key compromises [9].\nHowever, the protocol may be vulnerable if a parent and a child node in\nthe hierarchy are compromised.\nPrzydatek et al. proposed a secure information aggregation (SIA) frame-\nwork for sensor networks [10]. The framework consists of three node cate-\ngories: a home server, base station(s), and sensor nodes. A base station is a\nresources-enhanced node which is used as intermediary between the home\n" }, { "page_number": 486, "text": "Security in Wireless Sensor Networks\n■\n475\nserver and the sensor nodes, and it is also the candidate to perform the ag-\ngregation task. SIA assumes that each sensor has a unique identifier and\nshares a separate secret cryptographic key with both the home server and\nthe aggregator. The keys enable message authentication and encryption if\ndata confidentiality is required. Moreover, it further assumes that the home\nserver and base station can use a mechanism, such as μTESLA, to broad-\ncast authenticated messages. The proposed solution consists of three parts:\ncomputation of the result, committing to the collected data, and reporting\nthe aggregation result while proving the correctness of the result.\nIn the first part, the aggregator collects the data from sensors and locally\ncomputes the aggregation result. The aggregator can verify the authenticity\nof each sensor reading.\nIn the second part, the aggregator commits to the collected data. The\ncommitment to the input data ensures that the aggregator uses the data\nprovided by the sensors, and that the statement to be verified by the home\nserver about the correctness of computed results is meaningful. One ef-\nficient way of committing to the data is a Merkle hash-tree construction.\nIn this construction, all the data collected from the sensors is placed at the\nleaves of the tree. The aggregator then computes a binary hash tree starting\nfrom the leaf nodes. Each internal node in the hash tree is computed as\nthe hash value of the concatenation of its two child nodes. The root of the\ntree is called the commitment of the collected data. As the hash function\nin use is collision resistant, once the aggregator commits to the collected\nvalues, it cannot change any of the collected values. Figure 14.5 shows an\nexample of a Merkle hash tree.\nIn the third part, the aggregator and the home server engage in a pro-\ntocol in which the aggregator communicates the aggregation result and the\ncommitment to the server while proving to the server that the reported\nresults are correct using interactive proof protocols. Moreover, the authors\nalso presented efficient protocols for secure computation of the median\nand the average of the measurements, for the estimation of the network\nsize, and for finding the minimum and maximum sensor reading.\nDeng et al. proposed a collection of mechanisms for securing in-network\nprocessing (SINP) for WSNs [88]. Security mechanisms were proposed to\naddress the downstream requirement that sensor nodes authenticate com-\nmands disseminated from parent aggregators and the upstream requirement\nthat aggregators authenticate data produced by sensors before aggregating\nthat data. In the downstream stage, two techniques are involved: one way\nfunctions and μTESLA. The upstream stage requires that a pairwise key be\nshared between an aggregator and its sensor nodes.\nC¸ am et al. proposed an energy-efficient secure pattern-based data ag-\ngregation (ESPDA) protocol for wireless sensor networks in [89,90]. ESPDA\nis applicable for hierarchy-based sensor networks. In ESPDA, a clusterhead\nfirst requests sensor nodes to send the corresponding pattern code for the\n" }, { "page_number": 487, "text": "476\n■\nSecurity in Wireless Mesh Networks\nInternet\nHome with rooftop mesh router\nWired backbone connectivity\nGateway\nWireless link lower between mesh routers\nFigure 14.5\nMerkle hash tree used to commit to a set of values. The aggregator\nconstructs the Merkle hash tree over the sensor measurement m0, · · · , m7. To con-\nstruct the Merkle hash tree, the aggregator first hashes the measurements with a\ncryptographic hash function, e.g., v3,0 = H(m0), assuming that the size of the hash\nis smaller than the size of the data. Then, each internal value of the Merkle hash\ntree is derived from its two child nodes: vi, j = H(vi+1,2 j ∥vi+1,2 j+1). The Merkle\nhash tree is a commitment to all the leaf nodes. Once the aggregator commits to the\ncollected values, it cannot change any of the collected data. A verifier can authen-\nticate any value by verifying that the leaf value is used to derive the root node given\nthe authentic root node v0,0. For example, to authenticate the measurement m5, the\naggregator sends m5 along with v3,4, v2,3, v1,0, and m5 is authentic if the following\nequality holds: v0,0 = H(v1,0 ∥H(H(v3,4 ∥H(m5)) ∥v2,3)). (From Y. Wang, G. Atte-\nbury, and B. Ramamurthy, IEEE Communications Surveys and Tutorials, Vol. 8, no. 2,\npp. 2–23, 2006. With permission.)\nsensed data. If multiple sensor nodes send the same pattern code to the\nclusterhead, only one of them is permitted to send the data to the cluster-\nhead. ESPDA is secure because it does not require encrypted data to be\ndecrypted by clusterheads to perform data aggregation.\nFurther, C¸ am et al. introduced another secure differential data aggrega-\ntion (SDDA) scheme based on pattern codes [91]. SDDA prevents redun-\ndant data transmission from sensor nodes by implementing the following\nschemes: (1) SDDA transmits differential data rather than raw data, (2)\nSDDA performs data aggregation on pattern codes representing the main\ncharacteristics of sensed data, and (3) SDDA employs a sleep protocol to\n" }, { "page_number": 488, "text": "Security in Wireless Sensor Networks\n■\n477\ncoordinate the activation of sensing units in such a way that only one of\nthe sensor nodes capable of sensing the data is activated at a given time.\nIn the SDDA data transmission scheme, the raw data from sensor nodes is\ncompared to reference data with the difference data being transmitted. The\nreference data is obtained by taking the average of previously transmitted\ndata.\nDu et al. proposed a witness-based data aggregation (WDA) scheme for\nWSNs to assure the validation of the data sent from data fusion nodes to\nthe base station [92]. To prove the validity of the fusion result, the fusion\nnode has to provide proofs from several witnesses. A witness is one who\nalso conducts data fusion like a data fusion node, but does not forward its\nresult to the base station. Instead, each witness computes the MAC of the\nresult and then provides it to the data fusion node, which must forward the\nproofs to the base station.\nWagner studied secure data aggregation in sensor networks and pro-\nposed a mathematical framework for formally evaluating their security [93].\nIn [11] and [94], the authors proposed two data fusion schemes for the fil-\ntering of injected false data in sensor networks, which will be introduced\nin Section 14.8.\n14.7.2\nCiphertext-Based Secure Data Aggregation\nTwo ciphertext-based secure data aggregation schemes were proposed in\n[95] and [96]. The works in [95] and [96] are based on a particular encryp-\ntion transformation: a privacy homomorphism (PH). A privacy homomor-\nphism is an encryption transformation that allows direct computation on\nencrypted data. Let Q and R denote two rings, and let + denote addition\nand × denote multiplication on both. Let K be the key space. We denote\nan encryption transformation E : K × Q −→R and the corresponding\ndecryption transformation D : K × R −→Q. Given a, b ∈Q and k ∈K ,\nwe term\na + b = Dk(E k(a) + E k(b))\nadditively homomorphic and\na × b = Dk(E k(a) × E k(b))\nmultiplicatively homomorphic [12].\nThe proposed scheme, Concealed Data Aggregation (CDA), in [95] is\nbased on the PH proposed in [97]. Although the study in [98] has shown\nthat the proposed PH in [97] is unsecure against chosen plaintext attacks for\nsome parameter settings, the authors in [95] claimed that for the WSN data\naggregation scenario, the security level is still adequate and the proposed\n" }, { "page_number": 489, "text": "478\n■\nSecurity in Wireless Mesh Networks\nPH method in [97] can be employed for encryption. CDA can be used to\ncalculate SUM and AVERAGE in a hierarchical WSN. To calculate AVERAGE,\nan aggregator needs to know the number of sensor nodes n.\nCastelluccia et al. proposed a simple and provable secure additively ho-\nmomorphic stream cipher (HSC) that allows for the efficient aggregation of\nencrypted data [96]. The new cipher uses modular addition and is therefore\nvery well suited for CPU-constrained devices such as those in WSNs. The\naggregation based on this cipher can be used to efficiently compute statis-\ntical values such as the mean, variance, and standard deviation of sensed\ndata while achieving significant bandwidth gain.\n14.7.3\nOpen Research Issues\nData aggregation is essential for WSNs, and security is absolutely necessary\nto defend against compromised sensor nodes. Open research issues include\nthe following:\n■\nSeveral secure data aggregation protocols have been proposed.\nHowever, no comparisons have been conducted on these proto-\ncols. Further evaluations and comparisons are desirable to learn the\nperformance of these protocols. The performance matrices might\ninclude security, processing overhead, communication overhead,\nenergy consumption, and data compression rate.\n■\nNew data aggregation protocols need to be developed to address\nhigher scalability and higher reliability against aggregator and sensor\nnode cheating.\n14.8\nIntrusion Detection\nThe security mechanisms implemented in secure routing protocols and se-\ncure data aggregation protocols are configured ahead of time to inhibit an\nattacker from breaking the security of the network. These security mecha-\nnisms alone cannot ensure perfect security of a WSN. Because sensor nodes\ncan be compromised, it is easy for an adversary to inject false data into a\nWSN through the compromised nodes. Authentication and data encryption\nare not enough for ensuring data security. Another approach to protect\nWSNs involves mechanisms for detecting and reacting to intrusions.\nAn intrusion detection system (IDS) monitors a host or network for sus-\npicious activity patterns outside normal and expected behavior [5]. It is\nbased on the assumption that there exists a noticeable difference in the\nbehavior of an intruder and legitimate user in the network such that an\n" }, { "page_number": 490, "text": "Security in Wireless Sensor Networks\n■\n479\nIDS can match those pre-programmed or possibly learned rules. Based on\nthe analysis model used for analyzing the audit data to detect intrusions,\nintrusion detection systems in ad hoc networks are classified into rule-\nbased and anomaly-based systems. The rule-based intrusion detection sys-\ntems are used to detect known patterns of intrusions (e.g., [99] and [100])\nwhile anomaly-based systems are used to detect new or unknown intru-\nsions (e.g., [101] and [102]). A rule-based IDS has a low false-alarm rate\nwhen compared to an anomaly-based system, and an anomaly-based IDS\nhas a high intrusion detection rate in comparison to a rule-based system.\nHowever, WSNs are generally application-specific and lack basic infor-\nmation on topology, normal usage, expected communication patterns, etc.\nIt is impractical to pre-install some fixed patterns in sensors before they\nare deployed. Moreover, due to constraints in sensors, to learn and detect\nthese parameters after deployment is both time and energy consuming.\nThus, existing intrusion detection schemes in ad hoc networks may not be\nadapted to WSNs.\nThe research on intrusion detection in WSNs is still preliminary. Current\nresearch focuses on how to detect and eliminate injected false information.\nNote that compromised nodes can always inject false information into a\nsensor network. Thus, cooperation among sensors, especially neighboring\nnodes, is necessary to decide the validity of a report. In this section, we\ndiscuss the intrusion detection techniques in WSNs.\n14.8.1\nIntrusion Detection in WSNs\nZhu et al. proposed an interleaved hop-by-hop authentication (IHOP)\nscheme in [11]. IHOP guarantees that the base station will detect any in-\njected false data packets when no more than a certain number t of nodes\nare compromised. The sensor network is organized in a cluster-based hiera-\nrchy. Each clusterhead builds a route to the base station and each interme-\ndiate node has an upper associate node and a lower associate node that is\nt + 1 hops away.\nIHOP uses a number of shared keys:\n■\nEvery node shares a master secret key with the base station.\n■\nEach node knows its one-hop neighbors and has established a pair-\nwise key with each of them.\n■\nA node can establish a pairwise key with another node that is mul-\ntiple hops away if needed.\nFurther, IHOP also assumes that the base station has a mechanism to\nauthenticate broadcast messages, e.g., μTESLA.\nA clusterhead collects information from its members and sends a report\nto the base station only when at least t +1 sensors observe the same result.\n" }, { "page_number": 491, "text": "480\n■\nSecurity in Wireless Mesh Networks\nMeanwhile, a clusterhead also collects the MACs from detecting nodes.\nEach detecting node sends two MACs to the clusterhead: a MAC using the\nkey shared with the base station, referred to as the individual MAC, and a\nMAC using the key shared with its upper associate nodes, referred to as the\npairwise MAC. The clusterhead then compresses the t + 1 individual MACs\nby XORing them to reduce the size of a report. However, the pairwise\nMACs are not compressed for transmission. If they were, a node relaying\nthe message would not be able to extract the pairwise MACs of interest to it.\nThus, a legitimate report includes t + 1 pairwise MACs and a compressed\nMAC for the base station. When an intermediate node receives a report,\nit verifies the MAC of its lower associate node. If it fails, the report is\neliminated. Otherwise, it removes the MAC, generates a new MAC using its\nupper associate node pairwise key, and appends it to the report.\nIHOP ensures that the base station can detect false data packets when\nno more than t nodes are compromised. However, the paper does not show\nhow to select the parameter t for a sensor network.\nYe et al. proposed a statistical en-route filtering (SEF) mechanism that\ncan detect and drop false data in [94]. SEF uses a similar key assignment\nscheme as the basic random key scheme presented in [68]. There is a global\nkey pool and each sensor is pre-installed in a partition selected from the\npool. When a stimulus occurs in the fields, the sensors detecting this event\nelect one of the nodes as the center-of-stimulus (CoS), a node which col-\nlects and summarizes the detection results from all detecting nodes and\nproduces a synthesized report on behalf of the group. The CoS generates\nthe report and broadcasts it to all detecting nodes. If a detecting node agrees\nwith the report, it generates a MAC using a key in its partition and sends\nthe MAC to the CoS. The CoS reports the stimulus to the base station only\nif it receives adequate MACs. A legitimate report carries multiple MACs and\na single compromised node cannot fake all MACs. When an en-route node\nreceives the report, it verifies the correctness of the MACs probabilistically\nand drops those with invalid MACs immediately. Finally, if a report reaches\nthe base station, the base station checks all the MACs and filters out any\nremaining false reports that escaped the en-route filtering. When a stimu-\nlus appears, multiple nodes that detect it collaborate to process the signal\nand elect the CoS based on the sensing signal strength. The node with\nthe strongest signal stands out as the CoS. To reduce the communication\noverhead, SEF further uses a Bloom filter [103] to reduce MAC sizes. SEF\nis designed to protect against injected false information and cannot defend\nagainst selective forwarding attacks.\nDeng et al. proposed an intrusion-tolerant routing in wireless sensor\nnetworks (INSENS) in [104] and further evaluated its performance in [105].\nINSENS is a proactive routing protocol. The sensors collect local topology\ninformation and send this information back to the base station. The base\n" }, { "page_number": 492, "text": "Security in Wireless Sensor Networks\n■\n481\nstation generates a forwarding table based on the collected information\nand sends the routing table to the corresponding sensors. The base station\nis the central control point for calculating the routing table which relieves\nthe computation load of individual sensors. Protecting against intrusions\nfocuses on three attacks: DoS-type attacks, routing attacks, and select for-\nwarding attacks. To protect against DoS-type attacks, only the base station\nis allowed to broadcast to the entire network and individual sensors can\nonly send unicast messages. INSENS requires some broadcast authentica-\ntion scheme such as μTESLA. Although a compromised node may still alter\na valid message and broadcast that message to its neighbors, the damage\nis restricted to only nearby nodes and the downstream nodes. To pro-\ntect against routing attacks which propagate erroneous control packets, a\nsymmetric key is chosen for confidentiality and authentication. Further, to\nprotect against select forwarding attacks, data is sent to base stations along\ntwo separate paths which are calculated by the base stations in the route\ndiscovery step. However, INSENS is built on a table-based routing protocol,\nand as such depends on the base stations to collect all needed topology\ninformation to calculate the forwarding table for each individual sensor.\nThus, INSENS is not scalable in large sensor networks.\nWang et al. proposed a scheme to detect whether a node is faulty or\nmalicious with the collaboration of neighbor nodes [106]. In the proposed\nscheme, when a node suspects that one of its neighbors is faulty, it sends\nout messages to request the opinions on the behavior of this suspected\nnode from other neighbors of the suspect. After collecting the results, the\nnode analyzes the results to diagnose whether the suspect has a fault. The\nauthors formalized the problem as how to construct a dominating tree to\ncover all the neighbors of the suspect and further proposed two tree-based\npropagation collection protocols to construct a dominating tree and collect\ninformation via the tree structure.\n14.8.2\nOpen Research Issues\nIntrusion detection in WSNs is still largely open to research. Key research\nissues include the following:\n■\nDue to the constraints in WSNs, intrusion detection has many aspects\nnot of concern in other network types. The problem of intrusion\ndetection needs to be well defined in WSNs.\n■\nThe proposed IDS protocols in the literature focus on filtering in-\njected false information only [11,94,104]. These protocols need to\nbe improved to address scalability issues.\n" }, { "page_number": 493, "text": "482\n■\nSecurity in Wireless Mesh Networks\n14.9\nSecurity in WSNs: Future Directions\nWSNs are promising solutions for many applications, and security is often\na key concern. Although research efforts have been made on cryptogra-\nphy, key management, secure routing, secure data aggregation, and intru-\nsion detection in WSNs, there are still some challenges to be addressed.\nFirst, the selection of the appropriate cryptographic methods depends on\nthe processing capability of sensor nodes, indicating that there is no uni-\nfied solution for all sensor networks. Instead, the security mechanisms are\nhighly application-specific. Second, sensors are characterized by the con-\nstraints on energy, computation capability, memory, and communication\nbandwidth. The design of security services in WSNs must satisfy these con-\nstraints. Third, most of the current protocols assume that the sensor nodes\nand the base station are stationary. However, there may be situations, such\nas battlefield environments, where the base station and possibly the sen-\nsors need to be mobile. The mobility of sensor nodes has a great influence\non sensor network topology and thus raises many issues in secure routing\nprotocols. In particular, we identify some of the future directions in the\nstudy of security issues in WSNs as follows:\n■\nExploit the availability of private key operations on sensor nodes:\nRecent studies on public key cryptography show that public key\noperations may be practical in sensor nodes. However, private key\noperations are still too expensive to accomplish in a sensor node. As\npublic key cryptography can greatly ease the design of security in\nWSNs, improving the efficiency of private key operations on sensor\nnodes is highly desirable.\n■\nSecure routing protocols for mobile sensor networks: Mobility of\nsensor nodes has a great influence on sensor network topology and\nthus on the routing protocols. Mobility can be at the base station,\nsensor nodes, or both. Current protocols assume the sensor net-\nwork is stationary. New secure routing protocols for mobile sensor\nnetworks need to be developed.\n■\nContinuous stream security in WSNs: Current work on security in\nsensor networks focuses on discrete events such as temperature and\nhumidity. Continuous stream events such as video and images are\nnot discussed. Video and image sensors for WSNs might not be\nwidely available now, but will likely be in the future. Substantial\ndifferences in authentication and encryption exist between discrete\nevents and continuous events, indicating that there will be distinc-\ntions between continuous stream security and the current protocols\nin WSNs.\n■\nQoS and security: Performance is generally degraded with the ad-\ndition of security services in WSNs. Current studies on security in\n" }, { "page_number": 494, "text": "Security in Wireless Sensor Networks\n■\n483\nWSNs focus on individual topics such as key management, secure\nrouting, secure data aggregation, and intrusion detection. QoS and\nsecurity services need to be evaluated together in WSNs.\n14.10\nSummary\nAs WSNs grow in capability and are used more frequently, the need for\nsecurity in them becomes more apparent. However, the nature of nodes\nin WSNs gives rise to constraints such as limited energy, processing capa-\nbility, and storage capacity. These constraints make WSNs very different\nfrom traditional ad hoc wireless networks. As such, special protocols and\ntechniques have been developed for use in WSNs.\nWhile existing surveys in [12–15] discuss security in wireless networks,\nnone focus specifically on security in WSNs and the constraints unique to\nthem. In this chapter, we have surveyed the security issues in WSNs starting\nwith the attacks and countermeasures in each network layer followed by\nthe issues and solutions in cryptography, key management, secure routing,\nsecure data aggregation, and finally intrusion detection. Although the dis-\ncussed security services certainly add more computation, communication,\nand storage overhead in WSNs consuming more energy, they are highly\ndesirable and often required in real-world applications.\nAcknowledgments\nThis work is partially supported by NSF grant no. CCR-0311577.\nReferences\n[1]\nD. Estrin, L. Girod, G. Pottie, and M. Srivastava, Instrumenting the world\nwith wireless sensor networks, in Proceedings of the International Con-\nference on Acoustics, Speech and Signal Processing, Salt Lake City, Utah,\nMay 2001.\n[2]\nH. Chan and A. Perrig, Security and privacy in sensor networks, IEEE\nComputer Magazine, pp. 103–105, October 2003.\n[3]\nE. Shi and A. Perrig, Designing secure sensor networks, Wireless Commu-\nnication Magazine, vol. 11, no. 6, pp. 38–43, December 2004.\n[4]\nI. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, A survey\non sensor networks, IEEE Communications Magazine, vol. 40, no. 8,\npp. 102–114, August 2002.\n[5]\nA. D. Wood and J. A. Stankovic, Denial of service in sensor networks,\nIEEE Computer, vol. 35, no. 10, pp. 54–62, 2002.\n" }, { "page_number": 495, "text": "484\n■\nSecurity in Wireless Mesh Networks\n[6]\nN. Gura, A. Patel, A. Wander, H. Eberle, and S. C. Shantz, Comparing ellip-\ntic curve cryptography and RSA on 8-bit CPUs, in CHES ’04: Proceedings\nof Workshop on Cryptographic Hardware and Embedded Systems, August\n2004.\n[7]\nA. S. Wander, N. Gura, H. Eberle, V. Gupta, and S. C. Shantz, Energy anal-\nysis of public-key cryptography for wireless sensor networks, in PerCom\n’05: Proceedings of the 3rd IEEE International Conference on Pervasive\nComputing and Communication, March 2005.\n[8]\nA. Perrig, R. Szewczyk, V. Wen, D. Culler, and J. D. Tygar, SPINS: Secu-\nrity protocols for sensor networks, Wireless Networks, vol. 8, no. 5, pp.\n521–534, September 2002.\n[9]\nL. Hu and D. Evans, Secure aggregation for wireless networks, in Work-\nshop on Security and Assurance in Ad hoc Networks, 2003.\n[10]\nB. Przydatek, D. Song, and A. Perrig, SIA: secure information aggregation\nin sensor networks, in SenSys ’03: Proceedings of the 1st International Con-\nference on Embedded Networked Sensor Systems. New York, ACM Press,\n2003, pp. 255–265.\n[11]\nS. Zhu, S. Setia, S. Jajodia, and P. Ning, An interleaved hop-by-hop au-\nthentication scheme for filtering of injected false data in sensor networks,\nin Proceedings of IEEE Symposium on Security and Privacy, Oakland, CA,\nMay 2004, pp. 259–271.\n[12]\nR. L. Rivest, L. Adleman, and M. L. Dertouzos, On data banks and pri-\nvacy homomorphisms, in Foundations of Secure Computation (Workshop,\nGeorgia Institute of Technology, Atlanta, 1977), Academic, New York,\n1978, pp. 169–179.\n[13]\nF. Stajano and R. J. Anderson, The resurrecting duckling: Security is-\nsues for ad-hoc wireless networks, in Proceedings of the 7th Interna-\ntional Workshop on Security Protocols, London, Springer-Verlag, 2000,\npp. 172–194.\n[14]\nY.-C. Hu and A. Perrig, A survey of secure wireless ad hoc routing, IEEE\nSecurity & Privacy, special issue on Making Wireless Work, vol. 2, no. 3,\npp. 28–39, May/June 2004.\n[15]\nD. Djenouri, L. Khelladi, and N. Badache, A survey on security issues in\nmobile ad hoc and sensor networks, IEEE Communications and Surveys\nand Tutorials, vol. 7, no. 4, 2005.\n[16]\nP. Levis and D. Culler, Mate: A tiny virtual machine for sensor networks, in\nASPLOS-X: Proceedings of the 10th International Conference on Architec-\ntural Support for Programming Languages and Operating Systems, New\nYork, ACM Press, 2002, pp. 85–95.\n[17]\nEYES project, March 2002–February 2005 [online], available at http://www.\neyes.eu.org.\n[18]\nA. Savvides and M. B. Srivastava, A distributed computation platform for\nwireless embedded sensing, in ICCD ’02: Proceedings of the 2002 IEEE\nInternational Conference on Computer Design: VLSI in Computers and\nProcessors, Washington, DC, IEEE Computer Society, 2002, p. 220.\n[19]\nR. Kling, Intel research mote, in Network Embedded Systems Technology,\nWinter 2003 Retreat, January 15–17 2003.\n" }, { "page_number": 496, "text": "Security in Wireless Sensor Networks\n■\n485\n[20]\nJ. Hill, R. Szewczyk, A. Woo, S. Hollar, D. Culler, and K. Pister, System ar-\nchitecture directions for networked sensors, in ASPLOS-IX: Proceedings of\nthe 9th International Conference on Architectural Support for Program-\nming Languages and Operating Systems, New York, ACM Press, 2000,\npp. 93–104.\n[21]\nJ. Hill, R. Szewczyk, A. Woo, S. Hollar, D. Culler, and K. Pister, System\narchitecture directions for networked sensors, SIGOPS Operating Systems\nReview, vol. 34, no. 5, pp. 93–104, 2000.\n[22]\nS. Slijepcevic, M. Potkonjak, V. Tsiatsis, S. Zimbeck, and M. B. Srivas-\ntava, On communication security in wireless ad-hoc sensor networks,\nin Proceedings of 11th IEEE International Workshop on Enabling Tech-\nnologies: Infrastructure for Collaborative Enterprises (WETICE ’02), 2002,\npp. 139–144.\n[23]\nL. Yuan and G. Qu, Design space exploration for energy-efficient se-\ncure sensor networks, in ASAP ’02: IEEE International Conference on\nApplication-Specific Systems, Architectures, and Processors, July 2002, pp.\n88–100.\n[24]\nJ. Deng, R. Han, and S. Mishra, Enhancing Base Station Security in Wire-\nless Sensor Networks, Department of Computer Science, University of\nColorado, Technical report CU-CS-951-03, 2003.\n[25]\nB. Deb, S. Bhatnagar, and B. Nath, Information assurance in sensor net-\nworks, in WSNA ’03: Proceedings of the 2nd ACM International Confer-\nence on Wireless Sensor Networks and Applications, New York, ACM Press,\n2003, pp. 160–168.\n[26]\nC. Karlof and D. Wagner, Secure routing in wireless sensor networks:\nAttacks and countermeasures, in Proceedings of the 1st IEEE Interna-\ntional Workshop on Sensor Network Protocols and Applications, May 2003,\npp. 113–127.\n[27]\nJ. Newsome, R. Shi, D. Song, and A. Perrig, The sybil attack in sensor\nnetworks: Analysis and defenses, in IPSN ’04: Proceedings of IEEE Inter-\nnational Conference on Information Processing in Sensor Networks, April\n2004.\n[28]\nY.-C. Hu, A. Perrig, and D. B. Johnson, Packet leashes: A defense against\nwormhole attacks in wireless networks, in Proceedings of IEEE Infocomm\n2003, April 2003.\n[29]\nW. Diffie and M. E. Hellman, New directions in cryptography, IEEE Trans-\nactions on Information Theory, vol. 22, no. 6, pp. 644–654, November\n1976.\n[30]\nR. L. Rivest, A. Shamir, and L. Adleman, A method for obtaining digital\nsignatures and public-key cryptosystems, Communications of the ACM,\nvol. 26, no. 1, pp. 96–99, 1983.\n[31]\nD. W. Carman, P. S. Kruus, and B. J. Matt, Constraints and Approaches for\nDistributed Sensor Network Security, NAI Labs, Technical report 00-010,\n2000.\n[32]\nM. Brown, D. Cheung, D. Hankerson, J. L. Hernandez, M. Kirkup, and\nA. Menezes, PGP in constrained wireless devices, in Proceedings of 9th\nUSENIX Security Symposium, August 2000.\n" }, { "page_number": 497, "text": "486\n■\nSecurity in Wireless Mesh Networks\n[33]\nG. Gaubatz, J.-P. Kaps, and B. Sunar, Public key cryptography in sensor\nnetworks—Revisited, in ESAS ’04: 1st European Workshop on Security in\nAd-hoc and Sensor Networks, 2004.\n[34]\nM. O. Rabin, Digitalized Signatures and Public-Key Functions as In-\ntractable as Factorization, Cambridge, MA, Technical report, 1979.\n[35]\nJ. Hoffstein, J. Pipher, and J. H. Silverman, Ntru: A ring-based public\nkey cryptosystem, in ANTS-III: Proceedings of the 3rd International Sym-\nposium on Algorithmic Number Theory, London, Springer-Verlag, 1998,\npp. 267–288.\n[36]\nV. S. Miller, Use of elliptic curves in cryptography, in Lecture Notes in\nComputer Sciences; 218 on Advances in Cryptology—CRYPTO 85, New\nYork, Springer-Verlag, 1986, pp. 417–426.\n[37]\nN. Koblitz, Elliptic curve cryptosystems, Mathematics of Computation,\nvol. 48, pp. 203–209, 1987.\n[38]\nElliptic\nCurve\nCryptography,\nSECG\nStd.\nSEC1,\n2000,\navailable\nat\nwww.secg.org/collateral/sec1.pdf.\n[39]\nB. Kaliski, TWIRL and RSA Key Size, RSA Laboratories, Technical note,\nMay 2003.\n[40]\nRecommended Elliptic Curve Domain Parameters, SECG Std. SEC2, 2000,\navailable at www.secg.org/collateral/sec2.pdf.\n[41]\nD. Hankerson, A. Menezes, and S. Vanstone, Guide to Elliptic Curve Cryp-\ntography, New York, Springer-Verlag, 2004.\n[42]\nA. Freier, P. Karlton, and P. Kocher, The SSL Protocol, Version 3.0., avail-\nable at http://home.netscape.com/eng/ssl3/.\n[43]\nR. Watro, D. Kong, S. fen Cuti, C. Gardiner, C. Lynn, and P. Kruus, TinyPK:\nSecuring sensor networks with public key technology, in SASN ’04: Pro-\nceedings of the 2nd ACM Workshop on Security of Ad hoc and Sensor\nNetworks, New York, ACM Press, 2004, pp. 59–64.\n[44]\nA. Liu and P. Ning, TinyECC: Elliptic Curve Cryptography for Sen-\nsor Networks (version 0.1), September 2005, [online], available at\nhttp://discovery.csc.ncsu.edu/software/TinyECC/.\n[45]\nD. J. Malan, M. Welsh, and M. D. Smith, A public-key infrastructure\nfor key distribution in TinyOS based on elliptic curve cryptography,\nin Proceedings of the 1st IEEE International Conference on Sensor and\nAd Hoc Communications and Networks, Santa Clara, California, October\n2004.\n[46]\nC. Karlof, N. Sastry, and D. Wagner, Tinysec: A link layer security archi-\ntecture for wireless sensor networks, in SenSys ’04: Proceedings of the 2nd\nInternational Conference on Embedded Networked Sensor Systems, New\nYork, ACM Press, 2004, pp. 162–175.\n[47]\nU.S. National Institute of Standards and Technology (NIST), SKIPJACK and\nKEA algorithm specifications, Federal Information Processing Standards\nPublication 185 (FIPS PUB 185), June 1998.\n[48]\nR. L. Rivest, The RC5 encryption algorithm, in Fast Software Encryption,\nB. Preneel, Ed., Springer, 1995, pp. 86–96.\n[49]\nD. Eastlake III and P. Jones, U.S. Secure Hash Algorithm 1 (SHA1), RFC\n3174 (Informational), September 2001.\n" }, { "page_number": 498, "text": "Security in Wireless Sensor Networks\n■\n487\n[50]\nJ. Daemen and V. Rijmen, AES proposal: Rijndael, in Proceedings of 1st\nAES Conference, August 1998.\n[51]\nA. J. Menezes, S. A. Vanstone, and P. C. V. Oorschot, Handbook of Applied\nCryptography, CRC Press, Boca Raton, FL, 1996.\n[52]\nR. L. Rivest, The MD5 Message-Digest Algorithm, RFC 1321, April\n1992.\n[53]\nP. Ganesan, R. Venugopalan, P. Peddabachagari, A. Dean, F. Mueller, and\nM. Sichitiu, Analyzing and modeling encryption overhead for sensor net-\nwork nodes, in WSNA ’03: Proceedings of the 2nd ACM International\nConference on Wireless Sensor Networks and Applications, New York, ACM\nPress, 2003, pp. 151–159.\n[54]\nY. W. Law, S. Dulman, S. Etalle, and P. H. Hartel, Assessing Security-Critical\nEnergy-Efficient Sensor Networks, in Proceedings of 18th IFIP TC11 Inter-\nnational Conference on Information Security, Security and Privacy in the\nAge of Uncertainty (SEC), Athens, Greece: Kluwer Academic Publishers,\nBoston, May 2003, pp. 459–463.\n[55]\nD. J. Wheeler and R. M. Needham, TEA, a tiny encryption algorithm, in\nProceedings of Fast Software Encryption: Second International Workshop,\nLecture Notes in Computer Science, B. Preneel, Ed., vol. 1008, December\n14–16 1994.\n[56]\nR. L. Rivest, M. J. B. Robshaw, R. Sidney, and Y. L. Yin, The RC6 Block\nCipher, submitted to NIST as a candidate for the AES.\n[57]\nM. Matsui, New block encryption algorithm misty, in Proceedings of 4th\nInternational Workshop of Fast Software Encryption, E. Biham, Ed., LNCS\n1267, Springer-Verlag, 1997, pp. 54–68.\n[58]\nETSI/SAGE, Specification of the 3GPP Confidentiality and Integrity Algo-\nrithms Document 2: KASUMI Specification, December 1999.\n[59]\nK. Aoki, T. Ichikawa, M. Matsui, S. Moriai, J. Nakajima, and T. Tokita,\nSpecification of Camellia—A 128-Bit Block Cipher, Specification Version\n2.0, Nippon Telegraph and Telephone Corporation and Mitsubishi Electric\nCorporation, 2001.\n[60]\nY. W. Law, J. M. Doumen, and P. H. Hartel, Benchmarking block ciphers\nfor wireless sensor networks (extended abstract), in 1st IEEE International\nConference on Mobile Ad-hoc and Sensor Systems, IEEE Computer Society\nPress, October 2004.\n[61]\nS. A. Camtepe and B. Yener, Key distribution mechanisms for wireless sen-\nsor networks: A survey, Computer Science Department at RPI, Technical\nreport TR-05-07, 2005.\n[62]\nR. D. Pietro, L. V. Mancini, Y. W. Law, S. Etalle, and P. J. M. Havinga, LKHW:\nA directed diffusion-based secure multicast scheme for wireless sensor\nnetworks, in ICPPW ’03: Proceedings of the 32nd International Conference\non Parallel Processing Workshops, IEEE Computer Society Press, 2003,\npp. 397–406.\n[63]\nS. Zhu, S. Setia, and S. Jajodia, LEAP: efficient security mechanisms for\nlarge-scale distributed sensor networks, in CCS ’03: Proceedings of the\n10th ACM Conference on Computer and Communications Security, New\nYork, ACM Press, 2003, pp. 62–72.\n" }, { "page_number": 499, "text": "488\n■\nSecurity in Wireless Mesh Networks\n[64]\nB. Lai, S. Kim, and I. Verbauwhede, Scalable session key construction\nprotocols for wireless sensor networks, in IEEE Workshop on Large Scale\nReal Time and Embedded Systems, 2002.\n[65]\nS. A. Cametepe and B. Yener, Combinatorial design of key distribution\nmechanisms for wireless sensor networks, in Proceedings of 9th European\nSymposium on Research Computer Security, 2004.\n[66]\nJ. Lee and D. R. Stinson, Deterministic key predistribution schemes for\ndistributed sensor networks, in Proceedings of Selected Areas in Cryptog-\nraphy, 2004, pp. 294–307.\n[67]\nJ. Lee and D. R. Stinson, A combinatorial approach to key predistribu-\ntion for distributed sensor networks, in Proceedings of the IEEE Wireless\nCommunication and Networking Conference, 2005.\n[68]\nL. Eschenauer and V. D. Gligor, A key-management scheme for distributed\nsensor networks, in CCS ’02: Proceedings of the 9th ACM Conference on\nComputer and Communications Security, New York, ACM Press, 2002,\npp. 41–47.\n[69]\nH. Chan, A. Perrig, and D. Song, Random key predistribution schemes\nfor sensor networks, in Proceedings of IEEE Symposium on Security and\nPrivacy, May 2003.\n[70]\nD. Liu and P. Ning, Establishing pairwise keys in distributed sensor\nnetworks, in CCS ’03: Proceedings of the 10th ACM Conference on\nComputer and Communications Security, New York, ACM Press, 2003,\npp. 52–61.\n[71]\nR. D. Pietro, L. V. Mancini, and A. Mei, Random key-assignment for secure\nwireless sensor networks, in SASN ’03: Proceedings of the 1st ACM Work-\nshop on Security of Ad hoc and Sensor Networks, New York, ACM Press,\n2003, pp. 62–71.\n[72]\nW. Du, J. Deng, Y. S. Han, and P. K. Varshney, A pairwise key pre-\ndistribution scheme for wireless sensor networks, in CCS ’03: Proceedings\nof the 10th ACM Conference on Computer and Communications Security,\nNew York, ACM Press, 2003, pp. 42–51.\n[73]\nW. Du, J. Deng, Y. S. Han, S. Chen, and P. K. Varshney, A key management\nscheme for wireless sensor networks using deployment knowledge, in\nProceedings of IEEE INFOCOM, Hong Kong, 2004, pp. 586–597.\n[74]\nD. D. Hwang, B. Lai, and I. Verbauwhede, Energy-memory-security\ntradeoffs in distributed sensor networks, in Proceedings of 3rd Inter-\nnational Conference on Ad-hoc Networks and Wireless, July 2004, pp.\n70–81.\n[75]\nC. Blundo, A. D. Santis, A. Herzberg, S. Kutten, U. Vaccaro, and\nM. Yung, Perfectly-secure key distribution for dynamic conferences, in\nCRYPTO ’92: Proceedings of the 12th Annual International Cryptology\nConference on Advances in Cryptology, London, Springer-Verlag, 1993,\npp. 471–486.\n[76]\nR. Blom, An optimal class of symmetric key generation systems, in Pro-\nceedings of the EUROCRYPT ’84 Workshop on Advances in Cryptology:\nTheory and Application of Cryptographic Techniques, New York, Springer-\nVerlag, 1985, pp. 335–338.\n" }, { "page_number": 500, "text": "Security in Wireless Sensor Networks\n■\n489\n[77]\nD. Liu and P. Ning, Location-based pairwise key establishments for static\nsensor networks, in Proceedings of the ACM Workshop on Security in\nAd hoc and Sensor Networks, October 2003.\n[78]\nJ. N. Al-Karaki and A. E. Kamal, Routing techniques in wireless sensor\nnetworks: A survey, IEEE Wireless Communications, vol. 11, no. 6, pp. 6–\n28, December 2004.\n[79]\nC. Intanagonwiwat, R. Govindan, and D. Estrin, Directed diffusion: A scal-\nable and robust communication paradigm for sensor networks, in Mobi-\nCom ’00: Proceedings of the 6th Annual International Conference on Mo-\nbile Computing and Networking, New York, ACM Press, 2000, pp. 56–67.\n[80]\nY. Yu, R. Govindan, and D. Estrin, Geographical and Energy Aware Rout-\ning: A Recursive Data Dissemination Protocol for Wireless Sensor Net-\nworks, UCLA Computer Science Department, Technical report UCLA/\nCSD-TR-01-0023, May 2001.\n[81]\nT. Leinm¨uller, E. Schoch, F. Kargl, and C. Maih¨ofer, Influence of falsified\nposition data on geographic ad-hoc routing, in 2nd European Workshop\non Security and Privacy in Ad hoc and Sensor Networks (ESAS 2005),\nLNCS, July 2005.\n[82]\nD. Liu and P. Ning, Efficient distribution of key chain commitments for\nbroadcast authentication in distributed sensor networks, in Proceedings\nof the 10th Annual Network and Distributed System Security Symposium,\nSan Diego, CA, February 2003, pp. 263–276.\n[83]\nD. Liu and P. Ning, Multi-level μTESLA: Broadcast authentication for dis-\ntributed sensor networks, Transactions on Embedded Computing Systems,\nvol. 3, no. 4, pp. 800–836, 2004.\n[84]\nD. Liu, P. Ning, S. Zhu, and S. Jajodia, Practical broadcast authentication\nin sensor networks, in MobiQuitous ’05: Proceedings of The 2nd Annual\nInternational Conference on Mobile and Ubiquitous Systems: Networking\nand Services, July 2005, pp. 118–129.\n[85]\nW. Du, R. Wang, and P. Ning, An efficient scheme for authenticating public\nkeys in sensor networks, in MobiHoc ’05: Proceedings of the 6th ACM\nInternational Symposium on Mobile Ad hoc Networking and Computing,\nNew York, ACM Press, 2005, pp. 58–67.\n[86]\nPublic-Key Infrastructure (X.509) (pkix) [online], available at http://www.\nietf.org/html.charters/pkix-charter.html.\n[87]\nN. Shrivastava, C. Buragohain, D. Agrawal, and S. Suri, Medians and be-\nyond: new aggregation techniques for sensor networks, in SenSys ’04:\nProceedings of the 2nd International Conference on Embedded Networked\nSensor Systems, New York, ACM Press, 2004, pp. 239–249.\n[88]\nJ. Deng, R. Han, and S. Mishra, Security support for in-network processing\nin wireless sensor networks, in SASN ’03: Proceedings of the 1st ACM\nWorkshop on Security of Ad hoc and Sensor Networks, New York, ACM\nPress, 2003, pp. 83–93.\n[89]\nH. C¸ am, D. Muthuavinashiappan, and P. Nair, ESPDA: Energy-efficient\nand secure pattern-based data aggregation for wireless sensor net-\nworks, in Proceedings of IEEE Sensors, Toronto, Canada, October 2003,\npp. 732–736.\n" }, { "page_number": 501, "text": "490\n■\nSecurity in Wireless Mesh Networks\n[90]\nH. C¸ am, D. Muthuavinashiappan, and P. Nair, Energy-efficient security\nprotocol for wireless sensor networks, in Proceedings of IEEE VTC Con-\nference, Orlando, Florida, October 2003, pp. 2981–2984.\n[91]\nH. C¸ am, S. Ozdemir, H. O. Sanli, and P. Nair, Sensor Network Operations,\nWiley, 2004, ch. Secure Differential Data Aggregation for Wireless Sensor\nNetworks.\n[92]\nW. Du, J. Deng, Y. S. Han, and P. K. Varshney, A witness-based approach\nfor data fusion assurance in wireless sensor networks, in GLOBECOM ’03:\nProceedings of IEEE Global Telecommunications Conference, San Fran-\ncisco, December 2003, pp. 1435–1439.\n[93]\nD. Wagner, Resilient aggregation in sensor networks, in SASN ’04: Proceed-\nings of the 2nd ACM workshop on Security of Ad hoc and Sensor Networks,\nNew York, ACM Press, 2004, pp. 78–87.\n[94]\nF. Ye, L. Z. Haiyun Luo, and Songwu Lu, Statistical en-route filtering of\ninjected false data in sensor networks, in Proceedings of IEEE INFOCOM,\nHong Kong, 2004.\n[95]\nJ. Girao, D. Westhoff, and M. Schneider, CDA: Concealed data aggrega-\ntion for reverse multicast traffic in wireless sensor networks, in ICC ’05:\nProceedings of IEEE International Conference on Communications, Seoul,\nKorea, May 2005.\n[96]\nC. Castelluccia, E. Mykletun, and G. Tsudik, Efficient aggregation of en-\ncrypted data in wireless sensor network, in Proceedings of ACM/IEEE Mo-\nbiquitous, San Diego, July 2005.\n[97]\nJ. Domingo-Ferrer, A provably secure additive and multiplicative privacy\nhomomorphism, Lecture Notes in Computer Science, vol. 2433, pp. 471–\n483, 2002.\n[98]\nD. Wagner, Cryptanalysis of an algebraic privacy homomorphism, in ISC\n’03: Proceedings of the 6th Information Security Conference, Bristol, UK,\nOctober 2003.\n[99]\nS. Marti, T. J. Giuli, K. Lai, and M. Baker, Mitigating routing misbehav-\nior in mobile ad hoc networks, in MobiCom ’00: Proceedings of the 6th\nAnnual International Conference on Mobile Computing and Networking,\nNew York, ACM Press, 2000, pp. 255–265.\n[100]\nY. Zhang, W. Lee, and Y.-A. Huang, Intrusion detection techniques for\nmobile wireless networks, Wireless Networks, vol. 9, no. 5, pp. 545–556,\n2003.\n[101]\nY. Huang, W. Fan, W. Lee, and P. S. Yu, Cross-feature analysis for de-\ntecting ad-hoc routing anomalies, in ICDCS ’03: Proceedings of the 23rd\nInternational Conference on Distributed Computing Systems, Providence,\nRI, May 2003.\n[102]\nY. Huang and W. Lee, Attack analysis and detection for ad hoc routing\nprotocols, in RAIS ’04: Proceedings of the 7th International Symposium on\nRecent Advances in Intrusion Detection, Sophia Antipolis, France, Septem-\nber 2004.\n[103]\nB. H. Bloom, Space/time trade-offs in hash coding with allowable errors,\nCommunications of the ACM, vol. 13, no. 7, pp. 422–426, 1970.\n" }, { "page_number": 502, "text": "Security in Wireless Sensor Networks\n■\n491\n[104]\nJ. Deng, R. Han, and S. Mishra, INSENS: Intrusion-Tolerant Routing in\nWireless Sensor Networks, Department of Computer Science, University\nof Colorado, Technical report CU CS-939-02, November 2002.\n[105]\nJ. Deng, R. Han, and S. Mishra, A performance evaluation of intrusion-\ntolerant routing in wireless sensor networks, in IPSN ’03: Proceedings of\nIEEE 2nd International Workshop on Information Processing in Sensor\nNetworks, Palo Alto, California, 2003, pp. 349–364.\n[106]\nG. Wang, W. Zhang, C. Cao, and T. L. Porta, On supporting distributed\ncollaboration in sensor networks, in Proceedings of MILCOM, 2003.\n[107]\nY. Wang, G. Attebury, and B. Ramamurthy, A survey of security issues in\nwireless sensor networks, in IEEE Communications Surveys and Tutorials,\nVol. 8, no. 2, pp. 2–23, 2006.\n" }, { "page_number": 503, "text": "" }, { "page_number": 504, "text": "Chapter 15\nKey Management\nin Wireless Sensor\nNetworks\nFalko Dressler\nContents\n15.1\nIntroduction ........................................................ 494\n15.2\nSensor Network Security Objectives ............................... 495\n15.3\nApplication Scenarios .............................................. 496\n15.4\nKey Management in Sensor Networks............................. 498\n15.4.1\nOverview to Key Management............................ 498\n15.4.2\nKey Management Issues .................................. 499\n15.4.3\nKey Pre-Distribution....................................... 500\n15.4.4\nProactive Key Distribution ................................ 502\n15.5\nSelected Key Management Schemes ............................... 503\n15.5.1\nBalanced Random Pre-Distribution\nin Homogeneous Networks............................... 503\n15.5.2\nUnbalanced Random Pre-Distribution\nin Heterogeneous Networks .............................. 505\n15.5.3\nState-Based Key Pre-Distribution Supporting\nBusy–Sleep Cycles......................................... 506\n15.5.4\nTree-Based Key Distribution .............................. 508\n15.6\nOpen Research Challenges......................................... 510\n15.7\nConclusion.......................................................... 511\nReferences................................................................. 512\n493\n" }, { "page_number": 505, "text": "494\n■\nSecurity in Wireless Mesh Networks\nWireless sensor networks and corresponding applications greatly benefit\nfrom the proliferation of energy-aware embedded systems. Various appli-\ncation scenarios have successfully shown that the usage of sensor network\ntechnology is applicable in different domains. At the same time, the need\nfor security solutions is rising. This includes mechanisms for secure man-\nagement and control, e.g., routing and software management, as well as\nfor data communication. Similarly, the demand for higher availability in-\ncluding the protection against attacks and misbehaving nodes emerged.\nSecurity architectures have been proposed to address these requirements.\nAll these solutions are based on cryptographic algorithms and appropri-\nate key management and key distribution solutions. The objective of this\nchapter is to provide an overview to state-of-the-art key management and\nkey distribution techniques. Additionally, a classification of key manage-\nment and key distribution solutions is provided, followed by an in-depth\nstudy of selected key distribution approaches. The chapter also includes\nan outlook to application scenarios and outlines the open issues for further\nresearch on key management and key exchange.\n15.1\nIntroduction\nWireless sensor networks (WSN) have become a major research domain\nin the communications community [1]. Besides other issues that have been\nstudied so far [2], energy consumption and security were identified to be the\nmost challenging problem spaces. These properties are influenced by the\nmassively distributed operating principle based on self-organization mech-\nanisms [3]. Similarly, the lifetime of sensor networks [4] depends strongly\non the operation mode, i.e., the used routing algorithms, the application\nbehavior, and, finally, the employed security methods.\nA survey of security issues in ad hoc and sensor networks can be found\nin [5]. Additional related work in the security area, focused on WSN, is\nsummarized in [6].\nThe primary requirements on a successful security architecture are avail-\nability, authentication, data confidentiality, integrity, and non-repudiation.\nMost of these objectives can be addressed using cryptographic hash func-\ntions and appropriate encryption schemes. In ad hoc and sensor networks,\nmany proposals were published concerning the use of security measures\nfor particular applications [5]. Security protocols such as SPINS [6] define\ncomplex architectures to be used in a sensor network environment.\nMost of these proposals defer the problem of key management — one\nof the most sophisticated problems — to be solved elsewhere. Fortunately,\nseveral approaches seem to be adequate in this domain as already studied\nin ad hoc networks [7,8]. In this chapter, we discuss various key manage-\nment solutions for sensor networks and provide an overview to general key\n" }, { "page_number": 506, "text": "Key Management in Wireless Sensor Networks\n■\n495\npre-distribution and proactive key exchange solutions. This survey also pro-\nvides a classification of key management solutions for wireless sensor net-\nworks and an outline of open research issues including efficient public-key\nencryption in sensor networks [9]. Further discussion on key management\nsolutions can be found in [10].\nBesides security architectures and special solutions for routing or key\nmanagement, the aggregation of encrypted data in WSN was discussed [11]\nas well as the integration of particular security layers for reliable and secured\ncommunication [12]. Finally, secure overlays were proposed to address the\nsecurity concerns in WSN [13].\nIn summary, it can be said that many promising proposals can be found\nin the literature that address the security objectives in sensor networks.\nNevertheless, most of these papers only outline the principles or use simu-\nlation environments for verification. Experimentation on real sensor nodes\nis necessary to analyze the behavior of proposed security architectures and\nto contribute to the sensor network security domain.\nAll approaches for enabling security in WSN are very scenario depen-\ndent. There are different requirements, for example, in an agriculture\nscenario [14] compared to a habitat monitoring scenario [15]. Other require-\nments appear in the operation and control domain. Sensor nodes must be\nreconfigured, calibrated, and reprogrammed [16]. Such operations are very\nsensible for possible attacks. Finally, it must be mentioned that we ignore\nthe problem of key management. Several solutions have been proposed\nthat address this issue, e.g., [17].\nThe rest of this chapter is organized as follows. Section 15.2 outlines\nthe major security objectives in sensor networks. Then, Section 15.3 dis-\ncusses application scenarios that strongly depend on security mechanisms,\nand therefore profit from efficient and secure key management. This is fol-\nlowed by an overview to key management solutions and mechanisms in\nSection 15.4. Selected key management schemes are presented in detail in\nSection 15.5. Research challenges and open issues in key management are\noutlined in Section 15.6. Finally, Section 15.7 concludes the chapter.\n15.2\nSensor Network Security Objectives\nIn this section, we summarize the security properties required by commu-\nnication networks focusing on the specific capabilities of sensor networks.\nThe necessary security services in sensor networks are not altogether differ-\nent from those of other networks [5]. The goal of these services is to protect\ninformation and resources from attacks and misbehavior. In the context of\nsensor network security, the following requirements must be ensured for\nan effective security architecture.\n" }, { "page_number": 507, "text": "496\n■\nSecurity in Wireless Mesh Networks\n■\nData confidentiality: Ensures that the transmitted data cannot be un-\nderstood by anyone other than the desired recipient. Concentrating\non sensor networks, it is commonly agreed that the level of neces-\nsary confidentiality grows with the concentration or aggregation of\nmultiple sensor measures. Confidentiality is typically enabled by\napplying either symmetric or asymmetric data encryption techniques.\nTherefore, keys must be exchanged before a transmission can occur.\n■\nMessage authentication: Data or message authentication is of par-\namount importance for many applications in sensor networks. Tech-\nnically, message authentication ensures the genuineness of received\nmessages. Also covered is data integrity (see below). Usually, cryp-\ntographic hash functions using appropriate key material are used to\nfulfill this objective. In summary, data authentication ensures that\nreceived messages were sent by the expected source and not modi-\nfied during the transmission.\n■\nData integrity: Ensures that the received data was not modified dur-\ning the transmission. In contrast to message authentication, there\nis no key material involved in processes to ensure data integrity.\nSimilar cryptographic hash functions can be applied in this context.\nLooking at the properties of sensor networks, data integrity alone is\nnot sufficient due to the inherent property of multi-hop sensor net-\nworks that any node can intercept messages, modify them (including\nthe computation of a new hash value), and transmit the modified\nmessages to the final destination.\nA detailed analysis of security solutions for WSN is out of the scope of\nthis discussion. More information on this topic can be found in [5,6,18]. In\nsummary, it can be said that cryptographic hash functions and encryption\nschemes can be employed to ensure the most prominent security objectives\nin sensor networks. A prerequisite for this is the exchange of key material.\nThis step must occur before any sensor data can be exchanged.\n15.3\nApplication Scenarios\nThe security objectives as outlined in the previous section must be con-\nsidered in various application scenarios for wireless sensor networks. In\nthis section, we summarize selected applications that need to be secured\nby means of network security solutions. Additionally, we discuss the need\nfor inherently integrating key management solutions into the security ap-\nproaches to validate the efficiency and performance.\nOne of the first applications of network security mechanisms was secure\nrouting in ad hoc and sensor networks [18,19]. In most routing protocols,\nrouters exchange information on the topology of the network to establish\n" }, { "page_number": 508, "text": "Key Management in Wireless Sensor Networks\n■\n497\nroutes between nodes. Such information could become a target for mali-\ncious adversaries who intend to bring the network down. There are two\nsources of threats to routing protocols. The first comes from external attack-\ners [20]. By injecting erroneous routing information, replaying old routing\ninformation, or distorting routing information, an attacker could success-\nfully partition a network or introduce excessive traffic load into the network\nby causing retransmission and inefficient routing. The second and also the\nmore severe kind of threat comes from compromised nodes, which ad-\nvertise incorrect routing information to other nodes. Detection of such in-\ncorrect information is difficult: merely requiring routing information to be\nsigned by each node would not work, because compromised nodes are\nable to generate valid signatures using their private keys. Several solutions\nhave been proposed [18,21] that all rely on an efficient key management, in-\ncluding the detection of compromised or malicious nodes, and appropriate\nrevocation mechanisms are strongly demanded.\nSimilarly, the data dissemination and data forwarding needs to be\nsecured. Proposals such as SPINS [6] address this issue. Key management\ntechniques become even more critical if data must be aggregated, modified,\nor pre-processed within the network [22,23]. This case was, for example,\ndiscussed by Castelluccia and co-workers in their study on efficient aggre-\ngation of encrypted data in wireless sensor networks [11]. In this case, every\nnode that receives a packet needs to share a key with the sender to pro-\ncess the message. Key management can easily become unserviceable if too\nmany keys need to be stored in each device or if too many nodes become\ninvolved in a single-hop message exchange. We discuss this issue later in\nSection 15.5. Higher-layer solutions also rely on efficient key management\nthat is assumed to support end-to-end communication as well in a reliable\nand secure fashion [12].\nIf software modules are distributed in a sensor network, it must be\nverified that no attacker will be able to compromise a single node and\ndistribute modified, i.e., infected, software modules. Software management\nsolutions for sensor nodes were discussed in several proposals [16,24,25].\nKey management solutions must provide the basis for secured incremental\nnetwork programming for wireless sensors [25].\nService discovery is a more generalized form of knowledge distribution.\nIf specific services should be announced and used in a dynamic way, it must\nbe ensured that the identity of the service provider is unambiguous and it\nhas not been compromised so far [26]. A case study for secure distributed\nservice directory for wireless sensor networks outlined the needs of key\nmanagement solutions [27]. In this context, a secure overlay for service-\ncentric sensor networks was proposed [13].\nLooking at middleware applications such as service discovery, coordi-\nnation issues must be considered. Some of the most interesting solutions in\nthe context of ad hoc and sensor networks address security issues, including\n" }, { "page_number": 509, "text": "498\n■\nSecurity in Wireless Mesh Networks\nkey management objectives as well as particular challenges that emerge in\nsuch massively distributed systems. For example, a distributed coordina-\ntion framework for wireless sensor and actor networks was proposed [28]\nas well as a cooperation technique for self-organizing mobile ad hoc net-\nworks [29].\n15.4\nKey Management in Sensor Networks\n15.4.1\nOverview to Key Management\nThe organization of key management techniques strongly depends on the\nselected cryptographic scheme. As mentioned above, we only consider\ncryptographic hash and encryption mechanisms. In this section, we focus\non symmetric schemes that rely on appropriate key exchange and key dis-\ntribution instead of key verification. In Section 15.6, Open Research Chal-\nlenges, we give an outlook to issues for key management and verification\nfor asymmetric operations.\nKey management includes several functionalities. The most prominent,\nand in several solutions the only one, is key distribution. Nevertheless, key\nmanagement is also responsible for issues such as key revocation and re-\nkeying. Additionally, it must ensure resiliency to sensor-node capture. All\nthese issues are outlined in Section 15.4.2. In this sub-section, we present\na general classification of key distribution and key exchange solutions.\nIn theory, key management can be addressed in three ways:\n1.\nKey pre-distribution\n2.\nProactive key distribution\n3.\nOn-demand key exchange\nTo date, the only practical option for the distribution of keys to sensor\nnodes in a large-scale sensor network would have to rely on key pre-\ndistribution [30]. Keys would have to be installed in sensor nodes to ac-\ncommodate secure connectivity between nodes. However, traditional key\npre-distribution offers two inadequate solutions: either a single mission key\nor a set of separate n −1 keys, each being pairwise privately shared with\nanother node, must be installed in every sensor node. These and more\nrecent solutions that rely on probabilistic schemes [31] or on deployment\ninformation [32] are discussed in Section 15.4.3.\nProactive key distribution stands for key exchange after the deploy-\nment of the sensor network, but before any data communication occurs.\nProactive solutions usually rely on central base stations that provide the\nnecessary key material. On the other hand and to provide more reliabil-\nity, probabilistic solutions have been proposed that reduce the necessary\nkeys to a minimum, but still cover secure communication paths between all\n" }, { "page_number": 510, "text": "Key Management in Wireless Sensor Networks\n■\n499\nnodes [33]. Some of the proactive key distribution mechanisms also require\nsome pre-deployment actions such as the computation and selection of\nkey rings to be stored in all nodes [30]. Finally, tree-based key distribution\nalgorithms belong to this domain such as [10,34]. More detailed information\non proactive solutions is provided in Section 15.4.4.\nFinally, on-demand key exchange mechanisms address the needs of\ntypical applications not to focus on previously exchanged key material,\nbut to set up security relations on demand [35]. Public key solutions can\nbe seen to be on-demand solutions as the verification step takes place\nafter the communication was initiated [36]. In general, there are only a\nfew approaches available that make use of public-key cryptography. The\nprimary reason is the strong resource limitations in sensor networks, e.g.,\nthe computational power or the available memory. Novel approaches that\ncounteract these limitations are still works in progress such as [9].\n15.4.2\nKey Management Issues\nIn this sub-section, we present the basic features of key management\nsolutions. All solutions for key management basically concentrate on key\ndistribution or key pre-distribution. Nevertheless, issues such as revocation\nand re-keying must be considered as well.\n■\nKey distribution: Key distribution is the basis of all key management\nschemes [30]. It can be solved either by key pre-distribution prior\nto deployment or proactive in a sensor network prior to any data\ncommunication. Key distribution is the main topic of this chapter\nand is outlined in the following sub-sections.\n■\nRevocation: When a sensor node is compromised, it is essential to\nbe able to revoke keys associated with this sensor node. This may\ninvolve a complete new key distribution in case of a single mission\nkey. Usually, only the according key rings need to be discarded and\nre-built. Revocation procedures rely on an agreement that defines\nwhich keys need to be discarded. In most schemes, a controller\nnode coordinates such a process. If there is no central controller\navailable, election algorithms are used to select a node that performs\nthe necessary tasks.\n■\nRe-keying: The lifetime of (particular) keys can be limited using\nexpiration times. Although such mechanisms are rarely used in sen-\nsor networks, the expiration of keys and the necessary re-keying\nis a fundamental function in key management solutions. Basically,\nre-keying is equivalent to a self-revocation of a key by a node. It\ninvolves all nodes that share the specific key. Re-keying schemes\nwere categorized into two classes: stateful and stateless [17].\n" }, { "page_number": 511, "text": "500\n■\nSecurity in Wireless Mesh Networks\n■\nResiliency to sensor-node capture: The unattended operation of sen-\nsor nodes in hostile areas raises the possibility of sensor-node cap-\nture. Although node capture is a general threat that affects all security\nmechanisms, key management solutions must be aware of such sit-\nuations and provide adequate mechanisms to counteract such cap-\ntures. Basically, similar mechanisms as for general key revocation\ncan be used in this case.\n15.4.3\nKey Pre-Distribution\nTraditional Internet-based key exchange and key distribution protocols re-\nquire an infrastructure providing trusted third parties. Such approaches are\nnot feasible for large-scale sensor networks because the network topol-\nogy is not known prior to deployment, the communication range is very\nlimited, and the networks are dynamic in terms of sleep cycles or even\nnode failures. Therefore, most key management approaches are based on\nkey pre-distribution. Keys would have to be installed in sensor nodes to\naccommodate secure connectivity between nodes. Figure 15.1 depicts well-\nknown key pre-distribution schemes. The intention of key pre-distribution\nis to make key material available during or before the deployment to min-\nimize subsequent cryptographic overhead for key generation. In the fol-\nlowing sub-section, the schemes are explained and discussed.\n■\nSingle mission key: This approach deals with a pre-installed key on\nall sensor nodes. Usually, this key cannot be changed, and lasts for\nthe whole lifetime of the network. Depending on the scenario, a\nsingle mission key might be a feasible approach considering a small\nnetwork that needs to perform an application with a limited run-\ntime. In any other case, such a solution is inadequate because the\ncapture of any single node may compromise the complete network.\nAdditionally, attacks can be initiated to recover the key using eaves-\ndropped packets. Because all nodes use the same key, an attacker\nKey pre-distribution\nSingle mission\nkey\nn-1 keys\nRandom pre-\ndistribution\nDeployment\nknowledge\nFigure 15.1\nOverview of key pre-distribution techniques.\n" }, { "page_number": 512, "text": "Key Management in Wireless Sensor Networks\n■\n501\nwill be able to collect enough data for such an attack in quite a short\ntime. The selective revocation is not possible in this scenario.\n■\nSet of n−1 keys: In contrast to the single mission key approach, the\npairwise private sharing of keys between every two sensor nodes\navoids the compromising of the entire sensor network upon node\ncapture because selective key revocation becomes possible. How-\never, this solution requires pre-distribution and storage of n−1 keys\nin each sensor node and n(n −1)/2 per sensor network. It was\nshown in [30] that this approach is impractical for sensor networks\nconsisting of more than 10,000 nodes, for both intrinsic and tech-\nnological reasons. First, pairwise private key sharing between any\ntwo sensor nodes would be unusable because direct node-to-node\ncommunication is achievable only in small node neighborhoods de-\nlimited by communication range and sensor density. Second, incre-\nmental addition and deletion as well as re-keying of sensor nodes\nwould become both expensive and complex as they would require\nmultiple keying messages to be broadcast networkwide to all nodes\nduring their non-sleep periods (i.e., one broadcast message for every\nadded/deleted node or re-key operation). Third, a dedicated RAM\nmemory for storing n −1 keys would push the on-chip, sensor-\nmemory limits for the foreseeable future, even if only short, 64-bit\nkeys are used and would complicate fast key erasure upon detec-\ntion of physical sensor tampering. More scalable approaches in this\ncontext were proposed in [30,37].\n■\nRandom pre-distribution: The overhead due to the storage require-\nments for n(n −1)/2 keys can, for example, be reduced using ran-\ndomized techniques. Instead of storing the whole key ring for all\nn × n communication relationships, only samples of the complete\nkey ring are stored in each sensor node. To simplify the deployment\nof the sensor network as well as to allow the adding of nodes at any\ntime without the necessity of key exchange procedures, probabilistic\nmethods can be used to choose part of the key ring for each sensor.\nSuch scenarios were investigated by several groups [30,31,38]. The\ncomplexity of such approaches does not lie in the key management,\nbut in the identification of paths through the network that represent\ntrusted chains. In such a chain, two neighboring nodes must share\nidentical keys out of their key ring samples. So the problem of key\ndistribution can be reduced to the problem of path finding or rout-\ning. Specific solutions using random subset assignment and grid\nassignment techniques were studied in [39].\n■\nPre-distribution using deployment knowledge: Finally, another ap-\nproach can be used to reduce the storage requirements known from\nthe set of n −1 key solutions, the use of state information. Such\nsolutions exploit the deployment knowledge, i.e., the state of the\n" }, { "page_number": 513, "text": "502\n■\nSecurity in Wireless Mesh Networks\nsensors, to avoid unnecessary key assignments and to reduce the\nnumber of required keys that each sensor node should carry. At\nthe same time, it is possible to support higher connectivity and bet-\nter resilience against node failures. In this context, state informa-\ntion means the classification of sensor node states into active and\nsleep [32,40]. Using this information, the efficiency of pure proba-\nbilistic schemes can be noticeably improved.\n15.4.4\nProactive Key Distribution\nIn contrast to key pre-deployment strategies, proactive key distribution\nschemes are based on dynamic key generation or key exchange algorithms,\nrespectively. Most of these approaches need to be initialized by a key pre-\ndeployment mechanism as described above. Afterward, keys can be gen-\nerated and replaced dynamically. It must be mentioned that the dynamics\nin proactive solutions are limited. Compared to on-demand algorithms that\ncan create new keys just in time with a forthcoming communication [35],\nproactive mechanisms need to be executed prior to any data communi-\ncation, i.e., before the key material might be needed. Figure 15.2 depicts\nan overview of typical proactive key distribution methodologies. In the\nfollowing, possible solutions for such schemes are discussed.\n■\nBase station approach: Bootstrapping any further secured communi-\ncation can be initiated by selected base stations. Considering typical\nsensor network architectures, base stations are used to provide con-\nnectivity between the sensor network and a fixed communication\ninfrastructure. Therefore, compromising the base station could ren-\nder the entire sensor network useless. Thus, the base stations are a\nProbabilistic \nkey sharing \nBase station \napproach \nBalanced \nUnbalanced \nTree-based \nProactive key\nexchange\nFigure 15.2\nProactive key management techniques.\n" }, { "page_number": 514, "text": "Key Management in Wireless Sensor Networks\n■\n503\nnecessary part of the trusted computing base [6]. A trust setup mim-\nics this, and so all sensor nodes intimately trust the base station: at\ncreation time, each node is given a master key, which is shared with\nthe base station. All other keys are derived from this key.\n■\nProbabilistic key sharing: Another solution space is again based on\nprobabilistic schemes. Initially, trust is created by the use of subsets\nof key rings. The subsets can be either balanced, i.e., each node is re-\nquired to store the same amount of keys [30]. This procedure results\nin a homogeneous distribution of both, keys and subsequent pro-\ncessing requirements, due to key management actions. Depending\non the topology of the sensor network and the communication re-\nlationships, e.g., arbitrary communication vs. base station solutions,\nthis approach can lead to unfair exhaustion of resources of single\nsensor nodes. Additionally, heterogeneity of sensor nodes cannot be\nexploited, e.g., if the network consists of small nodes with very lim-\nited resources and larger ones that are able to store huge amounts\nof keys. Unbalanced approaches have been discussed that promise\nto solve this problem [33].\n■\nTree-based key management: In many sensor network scenarios,\neither the communication can be compared to a tree with a single\nbase station or gateway at the root [9] or the deployment follows a\nhierarchical structure [10]. In both cases, the key management can\nbe adapted to the tree structure to reduce the number of keys that\nneed to be pre-distributed or proactively computed.\n15.5\nSelected Key Management Schemes\nIn this section, we provide more details on selected key management\nschemes. Again, we follow the classification presented in the previous\nsection. Many proposed solutions are constructed on top of each other.\nTherefore, we try to follow the chronological order as well. The first three\nmethods, i.e., balanced random pre-distribution, unbalanced random pre-\ndistribution, and state-based pre-distribution, can directly be compared in\nterms of p(λ), the probability that two sensors share at least one key after\nthe pre-distribution phase. This parameter is outlined in each sub-section.\nAfterward, tree-based key distribution is discussed.\n15.5.1\nBalanced Random Pre-Distribution in Homogeneous\nNetworks\nEschenauer and Gligor presented a scheme for key management in dis-\ntributed sensor networks using probabilistic key sharing and a simple pro-\ntocol for shared-key discovery and path-key establishment, and for key\n" }, { "page_number": 515, "text": "504\n■\nSecurity in Wireless Mesh Networks\nrevocation, re-keying, and incremental addition of nodes [30]. Here, we\ndiscuss the three phases key pre-distribution, shared-key discovery, and\npath-key establishment.\nThe key pre-distribution phase consists of five offline steps:\n1.\nGeneration of a large pool of P keys (e.g., 217–220 keys) and of\ntheir key identifiers\n2.\nRandom drawing of k keys out of P without replacement to estab-\nlish a key ring of a sensor\n3.\nLoading the key ring into the memory of each sensor node\n4.\nSaving key identifiers of a key ring and associated sensor identifier\non a trusted controller node\n5.\nFor each node, loading the ith controller node with the key shared\nwith that node\nThis procedure ensures that only a small number of keys need to be placed\non each sensor node’s key ring to ensure that any two sensor nodes share\nat least a key with a chosen probability.\nThe shared-key discovery phase takes place during the sensor network\ninitialization. where every node discovers its neighbors in the wireless com-\nmunication range with which it shares keys. The simplest way to discover\nneighboring nodes that share a key with a specific node is to broadcast, in\ncleartext, the list of identifiers of the keys on the local key ring. Therefore,\nthis phase establishes the topology of the sensor network as seen by the\nnetwork layer. A link between any two neighboring nodes exists if they\nshare a key. The other way around, if a link exists between two nodes,\nall communication between these nodes can be secured using appropriate\ncryptographic algorithms.\nThe path-key establishment phase finally assigns a path-key to selected\npairs of nodes that do not share a key, but are connected by two or more\nlinks at the end of the shared-key discovery phase.\nUsing random graph theory, Eschenauer and Gligor have shown that,\ngiven a pool of P keys and randomly choosing k keys for the key ring, the\nprobability p of sharing a key between any two nodes in a neighborhood\ncan be calculated as follows:\np = 1 −Pr [two nodes do not share any key]\n= 1 −((P −k)!)2\n(P −2k)!P!\n(15.1)\nIn [30], the following numerical example was depicted. Let us assume a\nsensor network consisting of n = 10,000 nodes and a desired probability\nof Pc = 0.99999 for obtaining an “almost certainly” connected network, and\na wireless communication range that allows the neighborhood connectivity\n" }, { "page_number": 516, "text": "Key Management in Wireless Sensor Networks\n■\n505\nof 40 nodes. Then k = 250 out of P = 100,000 keys must be stored in each\nnode. If the connectivity increases to 60, only 200 keys are needed.\n15.5.2\nUnbalanced Random Pre-Distribution\nin Heterogeneous Networks\nTraynor and co-workers demonstrated that a probabilistic unbalanced dis-\ntribution of keys throughout the network that leverages the existence of a\nsmall percentage of more capable sensor nodes can not only provide an\nequal level of security, but also reduce the consequences of node com-\npromise. They demonstrated the effectiveness of this approach on small\nnetworks using a variety of trust models and then demonstrated the appli-\ncation of this method to very large systems [33].\nAs shown in the previous sub-section, random key pre-deployment in\nsensor networks has assumed very large random-graph arrangement such\nthat all neighbors within the transmission radius of a given node are reach-\nable. Communication between adjacent nodes is therefore limited only by\nkey matching. This model is not always realistic for a number of reasons.\nIn the unbalanced case, the network now consists of a mix of nodes with\ndifferent capabilities and missions. The sensing or Level 1 (L1) nodes are\nassumed to be very limited in terms of memory and processing capabil-\nity, and perform the task of data collection. Level 2 (L2) nodes have more\nmemory and processing ability. These nodes are equipped with additional\nkeys, and take on the role of routers and gateways between networks.\nAgain, the connectivity must be analyzed. In the following, n is the\nnumber of L1 nodes in a neighborhood, and g is the number of L2 nodes\nin a neighborhood, where applicable. The scheme for the unbalanced dis-\ntribution of keys throughout a wireless sensor network builds upon the\npreviously described balanced approach of Eschenauer and Gligor. Given\nthe same generated key pool of size P, we store a key ring of size k keys\nin each sensor (L1) node, and a key ring of size m keys in each L2 node,\nwhere m ≫k. Then, the probability of an L2 and L1 having at least one\nkey in common can be calculated as follows:\np = 1 −Pr [two nodes do not share any key]\n= 1 −(P −k)!(P −m)!\n(P −m −k)!P!\n(15.2)\nTraynor and co-workers demonstrated that their unbalanced approach has\nsimilar security capabilities as the balanced case. In a simulation, they have\nproven that a key ring of 328 keys (considering 40 neighboring nodes) is\ncomparable to 5 L2-nodes with 711 keys and 35 L1-nodes with 30 keys,\nrespectively. Therefore, they achieved a noticeable reduction of the load of\n" }, { "page_number": 517, "text": "506\n■\nSecurity in Wireless Mesh Networks\ntypical sensor nodes by exploiting heterogeneous sensor network environ-\nments. Additionally, the unbalanced scheme not only reduces the number\nof transmissions necessary to establish session-keys, but also reduces the\neffects of both single and multiple node captures. Lastly, the unbalanced\nscheme allows for even the most memory constrained platforms, from sen-\nsor nodes to RFID tags, to hold enough keys to establish secure connections\nfor communication.\n15.5.3\nState-Based Key Pre-Distribution Supporting\nBusy–Sleep Cycles\nLocation information can be facilitated as deployment knowledge for im-\nprovement of the previously discussed key pre-distribution schemes. If two\nsensor nodes are closely located to each other, they have very low prob-\nability to be in active-state at the same time. Therefore, unnecessary key\nassignments can be eliminated because keys shared only between such\nclosely located nodes may be hardly used. In [32,40], Park and co-workers\npropose a random key pre-distribution scheme that exploits new deploy-\nment knowledge, the state of the sensors, to avoid unnecessary key as-\nsignments and to reduce the number of required keys that each sensor\nnode must carry while supporting higher connectivity and better resilience\nagainst node captures.\nIn Figure 15.3, an example is shown for key assignments in a sensor net-\nwork. si and kj (with i = 1, 2, ... and j = 1, 2, ...) denote the sensor nodes\nand their pre-distributed keys, respectively. Let Ti denote the time-interval\nwhen sensor si is supposed to be in active-state with high probability. Two\nsensors, s1 and s2, are deployed closely, so they may share more keys as\ns1(k1, k2, k3, k4)\ns2\ns1\ns2 (k1, k3, k5, k6)\nActive node\nPassive node\nT1\nT2\nFigure 15.3\nTypical key assignments in sensor networks monitored at timeT1 andT2.\n" }, { "page_number": 518, "text": "Key Management in Wireless Sensor Networks\n■\n507\nproposed in [32]. Suppose that s1 and s2 have key set {k1, k2, k3, k4} and\n{k1, k3, k5, k6}, respectively. During T1, s1 and s2 are in active-state and\nsleep-state, respectively. Then, as time goes by, s1 and s2 transit their states\nto sleep and active, respectively. If s1 and s2 are in active-state at the same\ntime with very low probability, the shared key only between them, {k1, k3},\nmay be hardly used. Therefore, the key assignments of these keys to s1 and\ns2 are unnecessary.\nPark and co-workers used this idea to develop a state-based key man-\nagement scheme [40]. They assumed that sensor nodes are implemented\nto be in active-state at specific time-intervals with high probability and in\nother time-intervals the probability is relatively low. Then, sensor nodes\ncan be grouped by the time-intervals when they have high probabilities to\nbe in active-state. For instance, if sensor s1 has high probability to be in\nactive-state at time-interval T1, it may be grouped within the first group.\nUsing these assumptions, the active-state group (ASG) can be defined as\nthe group of sensor nodes with high probability to be in active-state at\nthe same time interval. The calculation of the active-probability is depicted\nin [40].\nFor key distribution, Park et al. use two key pools:\n1.\nGlobal key pool (GlP): A GlP S is a pool of random symmetric keys,\nfrom which a group key pool is generated. The cardinality of S is\nequal to |S|.\n2.\nGroup key pool (GrP): A GrP Si is a subset of GlP S for ith group,\nfrom which a key ring is generated. The cardinality of Si is equal\nto |SG|.\nThese pools are used for the key pre-distribution phase. Assuming L groups\ndefined during the modeling of the ASG, the key server generates a large\nGlP S and divides it into L GrPs Si for each ASG Gi. The purpose of setting\nup the GrP is to allow the time-neighbor ASGs to share more keys. After\ncompleting the GrP setup, for each sensor node j in ASG Gi, a randomly\nselected key ring R j,i from its corresponding GrP Si is loaded into the\nmemory of the sensors. For the assignment, an overlapping factor a is used\nthat determines a certain number of common keys between two nearby\ntime-interval groups. Because keys selected from the other groups are all\ndistinct, the sum of all the number of keys should be equal to |S|. Therefore,\n|SG| can be calculated as follows:\n|SG| =\n|S|\nL −aL + a\n(15.3)\nThe probability that two sensors share at least one common key can be\nexpressed as 1 −Pr [two nodes do not share any key]. Because the size of\nGrP is |SG|, the number of keys shared between two GrPs is λ|SG|, where\n" }, { "page_number": 519, "text": "508\n■\nSecurity in Wireless Mesh Networks\nis λ is 1, a, or 0. According to the value of λ, we should consider three\ncases for finding the required probability: two sensors come from same\ngroup (λ = 1), the neighbor two groups (λ = a), and the different groups\nwhich are not neighbors of each other (λ = 0). The same overlapping key\npool method used in [32] can be adopted. The first node selects i keys\nfrom the λ|SG| shared keys; it then selects the remaining R −i keys from\nthe non-shared keys. The second node selects R keys from the remaining\n|SG| −i keys from its GrP. Therefore, p(λ), the probability that two sensors\nshare at least one key when their GrPs have λ|SG| keys in common, can\nbe calculated as:\np (λ) = 1 −Pr [two nodes do not share any key]\n= 1 −\nmin(R,λ|SG|)\n\u0002\ni=0\n\u0003λ|SG|\ni\n\u0004 \u0003(1 −λ)|SG|\nR −i\n\u0004 \u0003|SG| −i\nR\n\u0004\n\u0003|SG|\nR\n\u00042\n(15.4)\nA detailed performance analysis of this approach is presented in [40]. In\nmany scenarios, this scheme offers a better performance compared to the\napproaches from Eschenauer and Gligor [30] and Du et al. [32].\n15.5.4\nTree-Based Key Distribution\nChen and Drissi contributed to the proactive key management by arranging\nthe sensor nodes in a hierarchical form [10]. They express the communi-\ncation in a sensor network in a well-structured way and provide several\napplication examples that support and confirm this approach. Given such\na hierarchical design of a sensor network as depicted in Figure 15.4, two\nforms of communication are necessary: between neighboring nodes at the\nA\nB\nC\nD\nF\nG\nH\nKG(A, B)\nKG(C, D)\nKG(F, G)\nK(A, F)\nK(F, H)\nLevel n – 1\nLevel n\nLevel n + 1\nK(D, G)\nFigure 15.4\nHierarchical or tree-based organization of sensors and the according\nkeys.\n" }, { "page_number": 520, "text": "Key Management in Wireless Sensor Networks\n■\n509\nsame level n (and the same group) and between sensors and their direct\nleaders in the next higher level n + 1.\nAppropriate keys must be distributed according to the communication\npaths in the network. Chen et al. propose the following scheme in which all\nnodes (except leaves and the root) are given four types of keys, namely, the\ngroup key (only one), the uplevel pairwise key (only one), the downlevel\ngroup key (only one), and the downlevel pairwise key (can be many).\nThese keys and their usage are described in the following. Hereby, we\nfollow the notation as used in Figure 15.4.\n■\nGroup key: The group key must be known by each group member to\ncommunication in the direct neighborhood, i.e., in the local group.\nExamples are nodes A and B, C and D, and F and G, respectively.\nA and B belong to the same group. Therefore, they must share the\nkey KG{A, B} for secure communication. This group key must also\nbe known by the direct group leader, i.e., node F in our example.\nThis knowledge is used for key management and command issues\ninstead of data communication.\n■\nDownlevel group key: The downlevel group key is the same key as\nthe group key described above. This key is only used for command\npurposes, e.g., key management issues for sensor node addition,\nreplacement, and deletion.\n■\nUplevel pairwise key: Communication between disjunctive groups\nmust occur via the network-inherent hierarchy, e.g., communication\nbetween A and C must use node F as a gateway. Therefore, each\nsensor node must share a private key with its uplevel group leader.\nExamples are pairwise keys K {A, F } between nodes A and F and\nK {F, H} between F and H.\n■\nDownlevel pairwise key: This key was is the same as the uplevel\npairwise key, but seen from the different angle.\nAs already mentioned, the communication paths follow the hierarchy as do\nthe key sharings. If node A wants to send a message to D, the following\ntransmissions will occur: A→F using K {A, F }, F→G using KG{F, G}, and\nG→D using K {D, G}.\nConsidering the performance of this approach, we examine the amount\nof keys necessary for communication and key management in such a hier-\narchical design. As described in [10], a network of n sensor nodes with a\ndepth of the tree of d (assuming a complete tree) results in logd n sensor\nnodes per group. Each leaf sensor only needs to store two keys; the root\nsensor needs to store approximately logd n + 1 keys. All the other nodes\nneed to store about logd n+3 keys. Therefore, the key storage requirement\nis O(logd n).\n" }, { "page_number": 521, "text": "510\n■\nSecurity in Wireless Mesh Networks\nA similar tree-based approach for secure key distribution is described by\nBla et al. [34]. In this work, the primary objective is on securely integrating\nnew nodes in an existing tree. Additionally, the hierarchical structure is not\nbased on a pre-defined setup, but on the real communication paths that\ncan be observed in the network.\n15.6\nOpen Research Challenges\nThe typical hardware and software constraints make it impractical to use the\nmajority of the current secure algorithms, which were designed for powerful\nworkstations. For example, the working memory of a sensor node is insuffi-\ncient even to hold the variables (of sufficient length to ensure security) that\nare required in asymmetric cryptographic algorithms (e.g., RSA and Diffie–\nHellman), let alone perform operations with them [6]. A particular challenge\nis broadcasting authenticated data to the entire sensor network. Current\nproposals for authenticated broadcast are impractical for sensor networks.\nFirst, most proposals rely on asymmetric digital signatures for the authen-\ntication, which are impractical for multiple reasons (e.g., long signatures\nwith high communication overhead of 50 to 1000 bytes per packet, very\nhigh overhead to create and verify the signature). The main problem of any\npublic key-based security system is to make each user’s public key available\nto others in such a way that its authenticity is verifiable. In mobile ad hoc\nnetworks, this problem becomes even more difficult to solve because of\nthe absence of centralized services and possible network partitions. More\nprecisely, two users willing to authenticate each other are likely to have\naccess only to a subset of nodes of the network (possibly those in their ge-\nographic neighborhood). Self-organized public key management is a first\napproach to address the security requirements in a scalable way [36]. On the\nother hand, cryptographic primitives are the fundamental building blocks\nof every secure protocol and the knowledge of algorithm usability is cru-\ncial for the design of new protocols for sensor networks. More acceptable\nencryption schemes using elliptic curve cryptography are proposed in [9].\nBroadcast authentication is another problem. Even previously proposed\npurely symmetric solutions for broadcast authentication are impractical:\nGennaro and Rohatgi’s initial work required over 1 KB of authentication in-\nformation per packet [41], and Rohatgi’s improved k-time signature scheme\nrequires over 300 bytes per packet [42]. Perrig et al. implemented the neces-\nsary primitives [6]. The available computational resources are usually very\nlimited and often not concerned security solutions. A typical performance\nevaluation must employ adequately calibrated simulation models [43]. In\nthis reference, measurements of typical sensor nodes are depicted that show\nthat even symmetrical cryptography has practical limitations in real sensor\nnetworks.\n" }, { "page_number": 522, "text": "Key Management in Wireless Sensor Networks\n■\n511\nA common characteristic of sensor networks is their severely limited\nenergy supply. Ultimately, the available energy determines that, for exam-\nple, base stations differ from nodes in having longer-lived energy supplies\nand having additional communications connections to outside networks.\nTo minimize the energy usage, a security sub-system should place mini-\nmal requirements on the processor, and add minimal information to each\nmessage transmitted. On the other hand, the limited lifespan of each node\nlimits the lifetime of usable keys. Given the severe hardware and energy\nconstraints, we must be careful in the choice of cryptographic primitives\nand the security protocols in the sensor networks.\nKey agreement is necessary based on scalable and efficient solutions.\nIn [44], three approaches to the problem of user-friendly key agreement\n(and mutual authentication) in settings where the users do not share any\nauthenticated information in advance were proposed. The first approach\nbelongs to the family of solutions requiring the users to compare strings of\nwords, whereas the other two approaches are based on radio channel spe-\ncific techniques, namely, distance-bounding and integrity-codes (I-codes).\nScalable key management with inherent self-configuration will allow the\ndeployment of even larger networks [45].\nLast but not least, group key management including group re-keying\nmechanisms for sensor networks are needed. Most existing group re-keying\nschemes are not suitable for sensor networks because they have large\noverhead and are not scalable. This problem was addressed by a family\nof pre-distribution and local collaboration-based group re-keying (PCGR)\nschemes [17]. These schemes are designed based on the ideas that future\ngroup keys can be preloaded to the sensor nodes before deployment, and\nneighbors can collaborate to protect and appropriately use the preloaded\nkeys.\nIn summary, the following research aspects and challenges for key man-\nagement solutions can be formulated:\n■\nEnergy-aware key management\n■\nPublic key management (key infrastructure)\n■\nFeasible public key cryptography\n■\nKey agreement mechanisms\n■\nGroup key management\n15.7\nConclusion\nSecurity issues in wireless sensor networks have been studied by various\ngroups to fulfill the raising demands of applications in this domain. In these\nworks, special requirements on security solutions have been identified that\nare correlated to the specific characteristics of sensor networks (strongly\n" }, { "page_number": 523, "text": "512\n■\nSecurity in Wireless Mesh Networks\nlimited resources in terms of processing and storage capacity, communi-\ncation bandwidth, and energy). Based on the results, many proposals for\nsecurity in WSNs are available that focus on routing, data aggregation, and\ncooperation issues. All of them rely on appropriate key management solu-\ntions that must be made available for sensor network installations.\nIn this chapter, we presented an overview to key management and key\ndistribution approaches for application in wireless sensor networks. We\nstarted with a first categorization of key management solutions in the area\nof WSN. Basically all proposals are based on efficient key pre-distribution\nor proactive key exchange supporting symmetric cryptographic techniques.\nThe different classes can be distinguished by the presumed knowledge\nabout network topology and routing mechanisms.\nBased on this classification, we described selected examples in detail\nto demonstrate the basic principles of the available solutions. We added a\nbrief discussion on the performance to each of these mechanism.\nBesides a few academic proposals and testbeds, asymmetric solutions\ncannot be found in sensor networks. There are two reasons for this ob-\nservation: first, asymmetric cryptographic operations cannot be efficiently\nused in small embedded systems and, second, to date there is no public\nkey infrastructure available for use in wireless sensor networks.\nFinally, we also provided a section outlining open issues and chal-\nlenges in the domain of security in WSN focusing on key management.\nThis roundup is intended to motivate further research work in this domain.\nReferences\n[1]\nI. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, A survey\non sensor networks, IEEE Communications Magazine, vol. 40, no. 8,\npp. 102–116, August 2002.\n[2]\nC.-Y. Chong and S. P. Kumar, Sensor networks: Evolution, opportunities,\nand challenges, Proceedings of the IEEE, vol. 91, no. 8, pp. 1247–1256,\nAugust 2003.\n[3]\nF. Dressler, Self-Organization in Ad Hoc Networks: Overview and Classi-\nfication, University of Erlangen, Dept. of Computer Science 7, Technical\nreport 02/06, March 2006.\n[4]\nF. Dressler and I. Dietrich, Lifetime analysis in heterogeneous sensor\nnetworks, in 9th EUROMICRO Conference on Digital System Design —\nArchitectures, Methods and Tools (DSD 2006), Dubrovnik, Croatia, August\n2006, pp. 606–613.\n[5]\nD. Djenouri and L. Khelladi, A survey of security issues in mobile ad hoc\nand sensor networks, IEEE Communication Surveys and Tutorials, vol. 7,\nno. 4, pp. 2–28, December 2005.\n" }, { "page_number": 524, "text": "Key Management in Wireless Sensor Networks\n■\n513\n[6]\nA. Perrig, R. Szewczyk, V. Wen, D. Culler, and J. D. Tygar, SPINS: Security\nprotocols for sensor networks, Wireless Networks, vol. 8, no. 5, pp. 521–534,\nSeptember 2002.\n[7]\nB. Wu, J. Wu, E. B. Fernandez, and S. Magliveras, Secure and efficient key\nmanagement in mobile ad hoc networks, in 1st International Workshop on\nSystems and Network Security (SNS 2005), April 2005.\n[8]\nJ.-P. Hubaux, L. Buttyan, and S. Capkun, Security, testbeds and applica-\ntions: The quest for security in mobile ad hoc networks, in ACM Interna-\ntional Symposium on Mobile and Ad Hoc Networks (ACM MobiHoc), October\n2001.\n[9]\nE.-O. Bla and M. Zitterbart, Towards acceptable public-key encryption in\nsensor networks, in the 2nd International Workshop on Ubiquitous Com-\nputing (ACM SIGMIS), May 2005.\n[10]\nX. Chen and J. Drissi, An efficient key management scheme in hierarchical\nsensor networks, in 2nd IEEE International Conference on Mobile Ad Hoc\nand Sensor Systems (IEEE MASS 2005): International Workshop on Wire-\nless and Sensor Networks Security (WSNS ’05), Washington, DC, November\n2005.\n[11]\nC. Castelluccia, E. Mykletun, and G. Tsudik, Efficient aggregation of en-\ncrypted data in wireless sensor networks, in Mobile and Ubiquitous Systems:\nNetworking and Services (MobiQuitous 2005), July 2005, pp. 109–117.\n[12]\nF. Dressler, Reliable and semi-reliable communication with authentication\nin mobile ad hoc networks, in 2nd IEEE International Conference on Mobile\nAd Hoc and Sensor Systems (IEEE MASS 2005): International Workshop\non Wireless and Sensor Networks Security (WSNS ’05), Washington, DC,\nNovember 2005, pp. 781–786.\n[13]\nH.-J. Hof, E.-O. Bla, and M. Zitterbart, Secure overlay for service centric\nwireless sensor networks, in 1st European Workshop on Security in Ad-Hoc\nand Sensor Networks (ESAS 2004), August 2004.\n[14]\nA. Baggio, Wireless sensor networks in precision agriculture, in ACM Work-\nshop on Real-World Wireless Sensor Networks (REALWSN 2005), Stockholm,\nSweden, June 2005.\n[15]\nA. Mainwaring, J. Polastre, R. Szewczyk, D. Culler, and J. Anderson, Wireless\nsensor networks for habitat monitoring, in 1st ACM Workshop on Wireless\nSensor Networks and Applications, Atlanta, September 2002.\n[16]\nG. Fuchs, S. Truchat, and F. Dressler, Distributed software management\nin sensor networks using profiling techniques, in 1st IEEE/ACM Interna-\ntional Conference on Communication System Software and Middleware\n(IEEE/ACM COMSWARE 2006): 1st International Workshop on Software\nfor Sensor Networks (SensorWare 2006), New Dehli, India, January 2006,\npp. 1–6.\n[17]\nW. Zhang and G. Cao, Group rekeying for filtering false data in sensor net-\nworks: A predistribution and local collaboration-based approach, in 24th\nIEEE Annual Joint Conference of the IEEE Computer and Communications\nSocieties (IEEE INFOCOM 2005), March 2005, pp. 503–514.\n[18]\nL. Zhou and Z. J. Haas, Securing ad hoc networks, IEEE Network, vol. 13,\nno. 6, November/December 1999.\n" }, { "page_number": 525, "text": "514\n■\nSecurity in Wireless Mesh Networks\n[19]\nB. R. Smith, S. Murphy, and J. J. Garcia-Luna-Aceves, Securing distance-\nvector routing protocols, in Symposium on Network and Distributed Systems\nSecurity, Los Alamitos, CA, February 1997, pp. 85–92.\n[20]\nC. Karlof and D. Wagner, Secure routing in wireless sensor networks: At-\ntacks and countermeasures, in Workshop on Sensor Network Protocols and\nApplications, 2003.\n[21]\nK. Sanzgiri, B. Dahill, B. N. Levine, C. Shields, and E. Belding-Royer, A\nsecure routing protocol for ad hoc networks, in International Conference\non Network Protocols (ICNP), November 2002.\n[22]\nL. Hu and D. Evans, Secure aggregation for wireless sensor networks, in\nWorkshop on Security and Assurance in Ad hoc Networks, 2003.\n[23]\nB. Przydatek, D. Song, and A. Perrig, SIA: Secure information aggregation\nin sensor networks, in ACM SenSys, 2003, pp. 255–265.\n[24]\nD. Culler, J. Hill, P. Buonadonna, R. Szewczyk, and A. Woo, A network-\ncentric approach to embedded software for tiny devices, in 1st Interna-\ntional Workshop on Embedded Software (EMSOFT 2001), Tahoe City, CA,\nOctober 2001.\n[25]\nJ. Jeong and D. Culler, Incremental network programming for wireless sen-\nsors, in 1st IEEE International Conference on Sensor and Ad hoc Commu-\nnications and Networks (IEEE SECON), June 2004.\n[26]\nF. Almenarez and C. Campo, SPDP: A secure service discovery protocol for\nad hoc networks, in 9th Open European Summer School and IFIP Workshop\non Next Generation Networks, Budapest, Hungary, 2003.\n[27]\nH.-J. Hof, E.-O. Bla, T. Fuhrmann, and M. Zitterbart, Design of a secure\ndistributed service directory for wireless sensor networks, in 1st European\nWorkshop on Wireless Sensor Networks, January 2004.\n[28]\nT. Melodia, D. Pompili, V. C. Gungor, and I. F. Akyildiz, A distributed co-\nordination framework for wireless sensor and actor networks, in 6th ACM\nInternational Symposium on Mobile Ad Hoc Networking and Computing\n(ACM Mobihoc 2005), Urbana-Champaign, Il, May 2005, pp. 99–110.\n[29]\nL. Buttyn and J.-P. Hubaux, Stimulating cooperation in self-organizing mo-\nbile ad hoc networks, Mobile Networks and Applications, vol. 8, no. 5,\npp. 579–592, October 2003.\n[30]\nL. Eschenauer and V. D. Gligor, A key-management scheme for distributed\nsensor networks, in 9th ACM Conference on Computer and Communication\nSecurity (ACM CCS), Washington, DC, November 2002.\n[31]\nH. Chan, A. Perrig, and D. Song, Random key management predistribution\nschemes for sensor networks, in IEEE Symposium on Research in Security\nand Privacy, 2003.\n[32]\nW. Du, J. Deng, Y. S. Han, S. Chen, and P. Varshney, A key management\nscheme for wireless sensor networks using deployment knowledge, in IEEE\nInfocom 2004, March 2004, pp. 586–597.\n[33]\nP. Traynor, H. Choi, G. Cao, S. Zhu, and T. L. Porta, Establishing pair-\nwise keys in heterogeneous sensor networks, in 25th IEEE Conference on\nComputer Communications (IEEE INFOCOM 2006), Barcelona, Spain, April\n2006.\n" }, { "page_number": 526, "text": "Key Management in Wireless Sensor Networks\n■\n515\n[34]\nE.-O. Bla, M. Conrad, and M. Zitterbart, A tree-based approach for secure\nkey distribution in wireless sensor networks, in The REALWSN, June 2005.\n[35]\nN. Asokan and P. Ginzboorg, Key agreement in ad hoc networks, Computer\nCommmunications, vol. 23, pp. 1627–1637, 2000.\n[36]\nS. Capkun, L. Buttyn, and J.-P. Hubaux, Self-organized public-key manage-\nment for mobile ad hoc networks, IEEE Transactions on Mobile Computing,\nvol. 2, no. 1, pp. 52–64, January 2003.\n[37]\nW. Du, J. Deng, Y. S. Han, and P. Varshney, A pairwise key predistribution\nscheme for wireless sensor networks, in 10th ACM Conference on Computer\nand Communications Security (CCS), October 2003, pp. 42–51.\n[38]\nS. Zhu, S. Xu, S. Setia, and S. Jajodia, Establishing pair-wise keys for se-\ncure communication in ad hoc networks: A probabilistic approach, in IEEE\nInternational Conference on Network Protocols (ICNP), November 2003.\n[39]\nD. Liu and P. Ning, Establishing pairwise keys in distributed sensor net-\nworks, in 10th ACM Conference on Computer and Communications Secu-\nrity, Washington DC, October 2003, pp. 52–61.\n[40]\nJ. Park, Z. Kim, and K. Kim, State-based key management scheme for\nwireless sensor networks, in 2nd IEEE International Conference on Mo-\nbile Ad Hoc and Sensor Systems (IEEE MASS 2005): International Workshop\non Wireless and Sensor Networks Security (WSNS ’05), Washington, DC,\nNovember 2005.\n[41]\nR. Gennaro and P. Rohatgi, How to sign digital streams, in Advances\nin Cryptology — Crypto ’97, vol. LNCS 1294, Berlin, Germany, 1997,\npp. 180–197.\n[42]\nP. Rohatgi, A compact and fast hybrid signature scheme for multicast packet\nauthentication, in 6th ACM Conference on Computer and Communication\nSecurity, November 1999.\n[43]\nM. Passing and F. Dressler, Experimental performance evaluation of crypto-\ngraphic algorithms on sensor nodes, in 3rd IEEE International Conference\non Mobile Ad Hoc and Sensor Systems (IEEE MASS 2006): 2nd IEEE Inter-\nnational Workshop on Wireless and Sensor Networks Security (WSNS ’06),\nVancouver, Canada, October 2006, pp. 882–887.\n[44]\nM. Cagalj, S. Capkun, and J.-P. Hubaux, Key agreement in peer-to-peer\nwireless networks, Proceedings of the IEEE (Special Issue on Cryptography\nand Security), vol. 94, no. 2, pp. 467–478, February 2006.\n[45]\nF. Liu and X. Cheng, A self-configured key establishment scheme for large-\nscale sensor networks, in 3rd IEEE International Conference on Mobile Ad\nHoc and Sensor Systems (IEEE MASS 2006), Vancouver, Canada, October\n2006, pp. 447–456.\n" }, { "page_number": 527, "text": "" }, { "page_number": 528, "text": "Index\nA\nAccess control, See Authorization; MAC\naddress security; Medium access\ncontrol (MAC) layer; specific\nmechanisms\nAccess Control Lists (ACLs), 367\nIEEE 802.15.4 networks, 417–418\nMAC spoofing vulnerability, 117\nZigBee security vulnerabilities,\n372–374\nAccess/One, 39, 77, 93–94\nAccess points (APs), 6, 48, 384\nauto-configurability, See\nAuto-configuration\nfriend nodes, 274–275\nhand-off mechanisms, 29\nhome networking, 10\nIEEE 802.11i authentication model,\n272–275\nintrusion detection issues, 153\nsecure routing approach, 178\nAcknowledgment (ACK), 21, 101, 449\nAdaptive Robust Tree (ART), 63–65\nAd hoc networks, 7–8, 47, See also Mobile\nad hoc networks\nenergy-aware protocols, 30\nenergy constraints, 7–8\nIEEE 802.11 mode, 35\ntransport layer protocols, 23–24\nAd hoc On-Demand Distance Vector\n(AODV)\nAODVSTAT, 160–161\nAOSR performance vs., 314–317\nhop count, 20–21\nIEEE 802.11s, 71, 175\nmessage formats and mutable fields,\n188–189\nrouting security issues, 179–182\nrushing attack vulnerability, 119\nSAODV, See Secure AODV\nsecure extensions, 191–193\nsecurity flaws of, 181–182\nsubnet communications, 180\ntrusted routing (TCAODV), 286–287\nAd-hoc On-demand Secure Routing\n(AOSR), 298, 306–310\nperformance evaluation, 313–317\nsecurity analysis, 310–313\nAd hoc QoS Routing (AQOR) protocol, 73\nAdministrative distances, 178, 336\nAdmission control, 74–75\nAdvanced Encryption Standard (AES), 354\nAES-CBC-MAC, 419\nAES-CCM, 368–369, 386, 419\ncounter mode (AES-CTR), 418\nIEEE 802.11i standard, 288\nIEEE 802.15.4 standard, 354–355,\n418–419\nZigBee, 369–370, 424\nAggregator nodes, 473, 475\nAlarms, 149, 159\nAlgebraic attacks, 364\nAlgorithmic key attacks, 389–390\nAnalog-to-digital converters (ADCs), 439\nAnomaly detection, 151–152, 155–156,\n165, 479\nAnonymity, 287\nAntenna technologies, 9, 16–17, 99\nAODV, See Ad hoc On-Demand Distance\nVector\nAODVSTAT, 160–161\n517\n" }, { "page_number": 529, "text": "518\n■\nSecurity in Wireless Mesh Networks\nAOSR, See Ad-hoc On-demand Secure\nRouting\nApplication layer, 101–102, 268, 440\nARAN, 130–131, 173\nAriadne, 130, 174–175\nArmenian Shuffle, 357\nAruba, 271\nAsymmetric key cryptography, 299, 301,\n326, 421, See also Public key\ncryptography; Simple Ad hoc Key\nManagement\nATCP, 24\nAttacks and vulnerabilities, 111, 115–125,\n146–147, 266–267, See also Security\nissues; specific network systems\nalgebraic, 364\nalgorithmic key, 389–390\nAODV vulnerabilities, 181–182\nattack signatures, 151, 165\nattacks on keys, 389–390\nback-door, 365\nblack hole, See Black hole attack\nbluesnarf, 365\nbrute-force, 389\ncontrol plane, 119–121, 163\ndata plane, 121–122\ndesynchronization, 450\ndictionary attacks, 389\nDoS, See Denial-of-service (DoS) attacks\necto-parasite and endo-parasite, 123–124\nflooding, 388, 449, 450\ngray hole, 121, 394\nhop integrity vulnerabilities, 198–199\ninternal and external, 113–114, 268,\n310, 441\nlaptop-class, 466\nlink layer, 115, 116, 446–447\nMAC layer, 115–118\nmalicious collisions, 446–447\nman-in-the-middle, 117–118, 135,\n365–366, 388, 391\nmemory exhaustion, 450\nmessage manipulation, 198–199, See also\nReplay attacks\nmote-class, 442\nmulti-radio multi-channel, 122–125\nnetwork layer, 119–122, 447–449,\n465–466\npartial matching and pre-computation,\n118, 139\nphysical layer, 115, 445–446\nreplay, See Replay attacks\nresource exhaustion, 447\nsession hijacking, 118, 135\nsinkholes, 120, 393–394, 448, 466\nsleep deprivation, 363, 374\nspoofing, See Spoofing attacks\nSybil, 121, 393, 448\ntime-memory tradeoff, 118\nwireless mesh LAN vulnerabilities,\n387–395, 405–406\nwireless sensor networks, 438, 443–450\nwormholes, See Wormhole attacks\nAuthenticated Routing for Ad hoc Network\n(ARAN), 130–131, 173\nAuthentication, 126, 267, 270–283, 467–469,\nSee also Key distribution; specific\nnetworks, protocols, or techniques\nAAA architectures for mesh networks,\n276–279, See also Authentication,\nauthorization, and accounting\nBluetooth mutual entity authentication,\n359–360\ncertification authorities, See Certificate\nauthority\ncluster level, 414–415\nComminus proposal (Tropos Network\nand Earthlink), 404–405\ndata packets, 275–276\nEAP, See Extensible Authentication\nProtocol\nfour-way handshake, 118, 133–134,\n137–139\nfriend nodes, 274–275\nhybrid architectures, 269\nIEEE 802.11i vulnerabilities, 135–139, See\nalso IEEE 802.11i\ninitial authentication protocol for hop\nintegrity, 198, 199, 202, 203–208\ninitialization vs. session phases, 269\nIntel proposal (IEEE 802.1X), 402–404\ninterleaved hop-by-hop (IHOP), 479–480\nlatency issues, 270\nLEAP, 421, 458–460\nlocation-based, 147\nMAC layer security mechanisms, 127–129\nμTESLA, 467–469, 475\nmulti-operator mesh networks, 282–283\nnetwork layer security mechanisms, 130\nnon-interactive key agreement protocol,\nSee Non-Interactive Key Agreement\nand Progression\noverhead, 292\npermutation vector, 128\nproblematic issues for wireless mesh\nnetworks, 139–140\nsecure routing requirements, 177\nsensor network security, 414–415, 441,\n467–469, 510\nsignatures, See Digital signatures\n" }, { "page_number": 530, "text": "Index\n■\n519\nsupplicant-authenticator dilemma, 396\ntoken-based reauthentication, 279–280\ntrust, See Trust\nwireless mesh LAN security, 386,\n396–399, 402–404, See also IEEE\n802.11i; Wireless mesh LANs\nAuthentication, authorization, and\naccounting (AAA), 262\narchitectures, 276–279\nComminus proposal for wireless mesh\nLANs, 405\nmobile clients and, 274, 292\nmulti-operator mesh networks,\n282–283\nservers, 132–133, 271, 277, 397\nAuthentication servers, wireless mesh LAN\nvulnerabilities, 396–397\nAuthenticator spoofing, 135–137\nAuthorization, 126, See also Authentication,\nauthorization, and accounting;\nspecific mechanisms\nEAP-TLS, 281\nIEEE 802.1X standard, 132–133\nsecure routing, 176–177\nsensor network security requirements,\n441\nwireless mesh LAN security, 396, 399\nAuto-configuration, 50, 292\ndeployment issues, 33\ndynamic address allocation, 329\nwireless LANs and mesh LANs, 68, 69\nAutomatic Repeat Request (ARR), 85\nAvailability issues\nadaptive support, 29–30\ndeployment, 33–34\nsensor networks, 411, 412, 441\nwireless mesh LANs, 399\nB\nBack-door attacks, 365\nBackhaul, defined, 48, 384\nBACnet IP, 40\nBACnet MSTP, 40\nBandwidth capacity, 14–15\nBase station (BS), 6–7, 9, 439\ncentralized key management, 458\nkey management scheme design\nconsiderations, 465\nproactive key distribution, 498–499,\n502–503\nsecure information aggregation\nframework, 474–475\nsmart antenna, 16\nSPINS, 325\ntrustworthiness assumption, 441\nwireless MAN architectures, 79–80\nBasic Service Set (BSS), 35, 68\nBattery lifetime, 352\nBattery power, 27, 30, 299, See also Power\nconsumption\nIEEE 802.15 mesh networking and, 36\nmesh PAN architecture and, 57\nsensor networks and, 412, 439\nsleep deprivation attack, 363, 374\nultra wide band and, 58\nZigBee vs. Bluetooth devices, 351, 353\nBeacons, 56\nIEEE 802.15.5 standard, 61\nsynchronization, 57\nZigBee MAC layer, 60\nBehavior-based detection, 151–152\nBit Error Rate (BER), 23\nBlack hole attack, 120–121, 140, 182, 267,\n394, 448\ncooperative attacks, 120–121\ndefenses, 131\nBlack list, 363\nBloom filter, 480\nBluejacking, 364\nBluesnarf attack, 365\nBluetooth, 350–352, 388\nbluejacking, 364\ncryptographic primitives, 354, 355–358\ndiscoverable modes, 352, 365\nDoS attacks, 363–364\nencryption algorithm E0, 355–356, 364\nhardware address, 352\nimplementation errors, 365\nkey agreement protocol, 358–361\nlocation privacy, 362\nman-in-the-middle attack vulnerability,\n365–366\nopen issues, 376\npiconet, 352\nPIN security, 362–363\nSAFER+ block cipher, 354, 355, 357, 365\nsecurity recommendations, 366\nsecurity weaknesses, 361–366\nspamming, 388\nunit key, 358, 362\nZigBee differences, 351, 376\nBluetooth Special Interest Group (SIG),\n350, 351–352\nBridging functions, 30\nBroadband and Wireless Network\n(BWN), 37\nBroadband wireless metropolitan area\nnetworks, See Metropolitan area\nnetworks\n" }, { "page_number": 531, "text": "520\n■\nSecurity in Wireless Mesh Networks\nBroadcast scheduling, 54\nBroadcast Session Key (BROSK), 459–460\nBrute-force attacks on keys, 389\nBusy Tone Multiple Access (BTMA), 16\nBWN-Mesh Testbed, 37\nC\nCalRadio-I, 37\nCamellia, 455\nCamera security, 390\nCATA, 17\nCatch, 266\nCBC-MAC Protocol (CCMP), 134–135\nAES-CBC-MAP under IEEE 802.15.4, 419\nencryption vulnerabilities, 139\nCCM, 368–369\nCCM *, 367, 368–369\nCCMP, See CBC-MAC Protocol\nCDMA, See Code Division Multiple Access\nCellular networks, 6\nhand-off mechanisms, 29\nmesh WAN architectures, 94\nCentralized scheduling, 80\nCertificate authority (CA), 299–300, 324,\n328, 329\nhop integrity protocol, 203–204\nnon-interactive key agreement\nprotocol, 300\nself-certified key cryptosystem, 301\nthreshold cryptography and, 317–319\nChannel capacity, 14–15\nChannel ecto-parasite attack (CEPA),\n123–124\nChannel reuse, 28\nChannel Switching Cost (CSC) metric, 22\nChannel Time Allocation (CTA), 55, 56\nChaska Wireless Internet Service\nProvider, 90\nChecksum mismatch, 446\nChittagong WiFi system, 93\nCipher Block Chaining MAC, See CBC-MAC\nProtocol\nCiphertext-based secure data aggregation,\n477–478\nCisco, 271\nClock synchronization requirements,\n173–175\nCluster-based networks, 409–414, See also\nIEEE 802.15.4\nIEEE 802.15.4 security services,\n415–419\nsecurity challenges and techniques,\n413–415\nCluster key establishment, 459, 463\nCode Division Multiple Access (CDMA), 28,\n97–100, 388\nCognitive radio, 99–100\nCollision Avoidance Time Allocation\n(CATA), 17\nCollisions, 446–447\nCollusion analysis, proposed privacy\npreserving solution, 247–255\nCombinatorial design theory (CDT), 460\nCommercial broadband access\ndeployment, 32\nComminus, 404–405\nCommunication overhead, 27–28, See also\nPower consumption\nCommunication range, 441\nCommunity networking, 5, 8, 10–11, See\nalso Metropolitan area networks;\nWiFi networks\nacademic research testbeds, 37\nComputation overhead, sensor nodes, 441\nCONFIDANT, 158–159, 266\nConfidentiality, 126, 287, 496, See also\nPrivacy preservation; specific security\nmechanisms or problems\ndata aggregation vs., 474\ndata and traffic, 229\neavesdropping attacks, See\nEavesdropping\nlightweight privacy preserving solution,\nSee Privacy preservation\nrouting security issues, 176\nsensor networks and, 414, 441\nContention Access Period (CAP), 56\nContention-based MAC protocols, 16–17\nControl plane attacks, 119–121, 163\nControl sub-frame, 36, 81\nCooperation issues, 265–266\nCooperative anomaly detection, 155–156\nCooperative black hole attack, 120–121\nCorpus Christi WiFi system, 91–92\nCritical nodes, 162–163\nCross-feature analysis, 131\nCross-layer routing design, 100\nCross-layer TCP optimization, 101\nCryptographically Generated Address\n(CGA), 330\nCSMA-CA algorithm, 410–411, 413\nD\nData aggregation\nciphertext-based, 477–478\nconfidentiality vs., 474\nIEEE 802.15.4 sensor networks, 412\nopen research issues, 478\n" }, { "page_number": 532, "text": "Index\n■\n521\nplaintext-based, 474–477\nprivacy homomorphism, 477–478\nsensor network security, 438,\n473–478, 497\nWDA, 477\nData confidentiality, See Confidentiality\nData integrity, 126, 267, 496\nhop integrity protocol, 197–199, 201–202,\n213–225\nintegrity check protocol, 198, 199, 202,\n214–225\nSDDA, 476–477\nsecure routing requirements, 177\nsensor network security, 415, 441\nSIA, 474–475\nZigBee vulnerabilities, 375\nData link layer, See Link layer\nData packets authentication, 275–276\nData plane attacks, 121–122, 163\ndata plane, 163\nData sub-frame, 36\nDead zones, 10, 51\nDefense-in-depth, 148\nDelayed signature verification, 334–337\nDempster–Shafer evidence theory, 164\nDenial-of-service (DoS) attacks, 113, 267,\n388, 391\nBluetooth vulnerabilities, 363–364\nbroadcast-based route discovery\nvulnerabilities, 392\ndistributed DDoS, 152, 391\nIDS responses, 152\ninfeasibility of preventing, 176–177\nmalicious collisions, 446–447\nmessage manipulation, 199\nresource exhaustion, 447\nsleep deprivation attack, 363\nwireless sensor network vulnerabilities,\n438, 445–451\nDeployment issues, 31–34\nauto-configurability, 33\ncommercial broadband access, 32\ncost, 32\ncoverage area, 33\ndeployment time, 33\nemergency operations, 31–32\nhome networking, 32\nincremental deployment, 32\nintegrating multiple network\ntechnologies, 33\nprotocol choice, 34\nservice availability, 33–34\nDeterministic key management approaches,\n421, 458–460\nDHCP discover messages, 390\nDIAMETER, 132, 271, 277–278\nDictionary attacks, 389\nDifferentiated Service (DiffServ), 53, 88\nDiffie–Hellman key agreement protocol,\n300, 404–405, 421, 451\nDigital signatures, 270\nforward-secure schemes, 188\nSAKM encoding, 337\nSAODV, 182, 184–186, 187, 190–193\nSAODV and delayed verification,\n334–337, 340–343\nsignature aggregation, 319\nDijkstra’s algorithm for path\ngeneration, 236\nDirected diffusion routing protocol,\n465–466, 470–471\nDirectional antennas, 9, 16–17, 99\nDirect sequence spread spectrum\n(DSSS), 59\nDisaster management applications, 12–13\nDiscoverable modes, 352, 365\nDistance-bounding, 511\nDistributed consensus protocol, 163\nDistributed denial-of-service (DDoS)\nattacks, 152, 391\nDistributed file storage, 11\nDistributed information sharing over WMN,\n101–102\nDistributed key management schemes, 458\nDistributed Laxity-based Priority\nScheduling, 18\nDistributed Packet Reservation Multiple\nAccess (D-PRMA), 17\nDistributed Priority Scheduling (DPS), 18\nDistributed Resolution Protocol (DRP), 62\nDistributed scheduling, 18, 80, 84\nDistributed Wireless Ordering Protocol\n(DWOP), 18\nDLPS, 18\nDOMINO, 266\nD-PRMA, 17\nDPS, 18\nDSDV-SQ, 173\nDSR, See Dynamic Source Routing\nDsynchronization attacks, 450\nDual-radio mode, 384\nDWOP, 18\nDynamic Source Routing (DSR), 20,\n22, 51\nsecure routing protocols, 173–174\nE\nE0 encryption algorithm, 355–356, 364\nEAP, See Extensible Authentication Protocol\n" }, { "page_number": 533, "text": "522\n■\nSecurity in Wireless Mesh Networks\nEarthlink, 404–405\nEavesdropping, 113, 116, 122, 387, 442\nECC, See Elliptic Curve Cryptography\nEcto-parasite attack, 123–124\nEdge routers, 8–9, 53\nElectronic beam forming, 16\nELFN, 24\nElliptic Curve Cryptography (ECC), 354,\n421, 452–454\nSAODV with delayed verification,\n342–343\nElliptic Curve Digital Signature Algorithm\n(ECDSA), 453\nEmergency applications, 12–13,\n31–32, 79\nEncryption key, Bluetooth, 358, 361\nEndo-parasite attack, 123\nEnd-to-end delay control, 75, 76,\n342–343\nEnd-to-end services, 125, 178, 182\nEnergy-aware protocols, 30\nEnergy efficiency of nodes, 27\nEnergy resource constraints, See Power\nconsumption\nEnhanced Distributed Control Function\n(EDCF), 89\nEnterprise networking applications, 51,\n66–67\nEntropy, 233–236\nError correction, 85, 447\nError message security, 173, 180–181,\n182, 186\nESPDA, 475–476\nEthereal, 150\nEthernet IP, 40\nETX, 21, 22, 51\nEvent-detection application, 413\nEvil twin attacks, 388\nExpected transmission count (ETX), 21,\n22, 51\nExpert system, 150\nExplicit Link Failure Notification (ELFN), 24\nExport authorization and routing\nsecurity, 176\nExtended Service Set (ESS), 35, 66, 382\nExtensible Authentication Protocol (EAP),\n132–133\nadaptive EAP-TLS and proxy chaining,\n281–282\nEAP-TLS, 279\nEAP-TLS over PANA, 280–281\nsupplicant-authenticator dilemma, 396\ntoken-based reauthentication, 279–280\nwireless mesh LAN security\nprotocols, 386\nF\nFAMA, 16\nFault tolerance, 166, 444\nFinite State Machine (FSM), 355\nFireTide, 40\nFirewalls, 153\nFixed channel, 20\nFlash memory, 441\nFlooding attacks, 388, 449, 450, See also\nDenial-of-service (DoS) attacks\nFloor Acquisition Multiple Access\n(FAMA), 16\nForgery, 388\nForward Error Correction (FEC), 86\nForward-secure signature schemes, 188\nFour-way handshake, 133–134, 272, 275,\n400\nvulnerabilities, 118, 137–139\nFrequency Division Multiplexing (FDM), 15\nFrequency-hopping spread spectrum\n(FHSS), 438, 446\nFrequency-selective fading channel, 15\nFriend nodes, 274–275\nFull function devices (FFDs), 57, 59\nG\nGame theoretic approach, 130\nGas meter reading, 92\nGated, 176\nGateway routers, 4, 5, 48, 112, See also\nRouting\nload balancing, 24–26\nprivacy preserving architecture, 232,\nSee also Privacy preservation\nGeographic and Energy Aware Routing\n(GEAR), 466\nGeorgia Tech, 37\nGlobal keys, 270\nGray hole attack, 121, 394\nGreedy nodes, 266, 395–396\nGreedy Perimeter Stateless Routing\nprotocol, 89\nGrid Computing paradigm, 292\nGroup cipher suite, 403\nGroup keys, 318, 419\nglobal key security issues, 270\nIEEE 802.11i, 400\nIEEE 802.15.4 sensor networks,\n421–422\nmesh WLAN challenges, 406\nopen issues for sensor networks, 511\ntree-based distribution, 509\nZigBee vulnerabilities, 373\n" }, { "page_number": 534, "text": "Index\n■\n523\nGroup Temporal Key (GTK), 133\nGumStix, 388\nH\nHand-off management, 28–29, 34, 52–53,\n85–86, 95–96, 98, 102, 278\nHash functions, 299, 424, 496\nAOSR for NIKAP, 311\nencryption algorithm overhead vs., 455\npacket authentication, 276\nSAKM message fields, 332\nsalt variations, 333\nSAODV, 182–184\nsecure routing protocols, 173\nZigBee Alliance specification (IEEE\n802.15.4), 424, 427\nHash message authentication code\n(HMAC), 424\nHash-tree, 475–476\nHealth care applications, 67\nHeartbeats, 53\nHello flood attacks, 449\nHome networking, 10, 51, See also Personal\narea networks\ndeployment issues, 32\nWLAN vs. wireless mesh LAN\napplications, 66, See also Wireless\nlocal area networks; Wireless\nmesh LANs\nHomomorphic stream cipher (HSC), 478\nHoneypots, 394\nHop count, 20–21, 51, 180–184, 187–188\nHop integrity protocol, 197–226\nAbstract Protocol Notation, 199–201\nconcept, 201\ninitial authentication protocol, 198, 199,\n202, 203–208\nintegrity check protocol, 198, 199, 202,\n214–225\nopen issues, 225\nrequirements of, 201\nsecret exchange protocol, 198, 199, 202,\n208–214\nsecurity threats, 198–199\nstrategic deployment, 225\nHop Reservation Multiple Access\n(HRMA), 18\nHospital applications, 67\nHost-based intrusion detection systems, 149\nHRMA, 18\nHybrid keying models, 421\nHybrid Wireless Mesh Protocol (HWMP),\n392, 395\nI\nICMP flood, 388\nID-based cryptography, 318\nIDEA, 455\nIdentity issues, 327–329\nIEEE 802.11, 6\nInter Access Point Protocol (IAPP), 278\nkey security vulnerabilities, 389–390\nmesh networking products, 39\nmulti-channel MAC vs., 18–19\nnetwork deployments/testbeds, 34–35,\n37–38\nIEEE 802.11a, 14, 35\nIEEE 802.11b, 14, 34, 37, 38, See also WiFi\nnetworks\nIEEE 802.11e, 35\nIEEE 802.11f, 278\nIEEE 802.11g, 35, 37, 38\nIEEE 802.11i, 118, 128, 383\nAES specification, 288\nauthentication model, 272–275\nComminus proposal (Tropos Network\nand Earthlink), 404–405\ncurrent security proposals, 401–406\nforwarding support, 274–275\nkey storage, 271\nMAC layer security, 132–135\nmanagement frame security, 397–398\nmulti-hop network vulnerabilities, 273\nopen issues, 140\nsecurity vulnerabilities, 135–139\nsession key storage, 271\nState-Based Key Hop (SBKH)\nprotocol, 288\nwireless mesh LAN security protocols,\n385–386\nIEEE 802.11s, 67–73, 175, 400–405\nIEEE 802.11w, 397\nIEEE 802.15, 6, 16, 35–36, 352, See also\nBluetooth\nIEEE 802.15.4, 59–60, 351, 352, 354–355,\n409–431, See also Wireless sensor\nnetworks; ZigBee\naddressing, 415–418\nAES specification, 354–355, 418–419\nbandwidth capacity, 411\ndata aggregation, 412\nIV (nonce) management, 372, 418\nkey management models, 419–423\nnetwork topologies, 410\npower consumption, 412\nreplay protection, 419\nsecurity challenges and techniques,\n413–415\n" }, { "page_number": 535, "text": "524\n■\nSecurity in Wireless Mesh Networks\nsecurity limitations, 423\nsecurity operations, 415–419\nZigBee security services for sensor\nnetworks, 423–431\nIEEE 802.15.5, 57, 60–65\nIEEE 802.16, 6, 36–37, 39, 78\nMAC layer, 81–85\nMIMO, 16\nmobility management, 85–86\nnetwork deployments/testbeds, 36–37\nQoS provisions, 53\nIEEE 802.16a, 36\nIEEE 802.16e, 36, 86, 94–96\nIEEE 802.1X, 132–133, 383, 402–404\nEAP variants, 279–282\nport control mechanisms, 395\nsecurity vulnerabilities, 135–137\nwireless mesh LAN security\nprotocols, 386\nIEEE 802.20, 6, 94, 95, 96–99\nIETF MANET Work Group, 382–383\nIHOP, 479–480\nIKA2, 421\nImpersonation, 148, 172, 181–182, 198, 268,\n271, 362, 388, 392, 405, See also\nMan-in-the-middle attacks; Spoofing\nattacks\nImport authorization and routing security,\n176–178, 334–335\nIncremental deployment, 32\nIndependent Basic Service Set (IBSS), 68\nIndex of load balance (ILB), 25\nIndustrial research, 38–39\nInformation theory and privacy\npreservation, 256–257, 289\nInfrastructure backbone networking, 8, 35\nInfrastructure wireless networks, 6\nInitial authentication protocol, 198, 199,\n202, 203–208, 269\nInitialization key, Bluetooth, 358–359\nInstruction Set Architecture (ISA), 455\nIntegrated Service (IntServ), 53\nIntegrity check protocol, 198, 199,\n214–225\nstrong protocol, 218–225\nweak protocol, 214–218\nIntegrity-codes (I-codes), 511\nIntegrity of data, See Data integrity\nIntel, 39, 402–404\nIntelligent transportation systems, 79\nIntel Mote, 441\nInter Access Point Protocol (IAPP), 278\nInterference, See Radio interference\nInterference-Aware Resource Usage\n(IRU), 22\nInterleaved hop-by-hop (IHOP)\nauthentication, 479–480\nInternal attacks and vulnerabilities,\n113–114, 268, 310, 441\nInternet access, 5, 9, 78–79\nInternet-based intrusion detection schemes,\n154–155\nInternet Key Exchange (IKE), 281\nInternet service provider (ISP), 78–79, 90\nIntrinsic quality of service, 53\nIntrusion detection, 130, 145–166, 267, 268,\nSee also Misbehavior detection\nalarms, 149, 159\nanomaly detection, 151, 155–156,\n165, 479\nattack signatures, 151\nCONFIDANT, 158–159\ncooperative anomaly detection, 155–156\ncritical nodes, 162–163\ncross-feature analysis, 131\ndefense-in-depth, 148\nDempster–Shafer evidence theory, 164\nevaluation issues, 165–166\nfalse positives and negatives, 149, 165\nfirewalls and, 153\nflow status messages and TIARA, 157\ngame theoretic approach, 130\ngoals of, 149\nhost-based and network-based\nmonitoring, 149–150\nIHOP authentication, 479–480\nInternet-based schemes, 154–155\nlimited resource usage, 164\nMAC spoofing and, 117\nmalcounts, 157–158\nmisuse detection, 150–151, 156,\n161, 165\nMobIDS, 159–160\nmobile agents, 160\nmobility issues, 153–154\nnetwork layer, 131–132\nopen research issues, 165, 481\npacket sniffers, 149–150\npathraters, 156–157\nproblematic issues, 139\nRESANE, 162\nresponses, 152\nSCAN, 163–164\nscheme for Internet environment,\n154–155\nspecial challenges for mesh networks,\n146, 152–154\nSTAT and AODVSTAT, 160–161\ntrust model, 161–162\nwatchdogs, 156–157\n" }, { "page_number": 536, "text": "Index\n■\n525\nWATCHERS, 154–155\nwireless sensor networks and, 478–481\nIntrusion prevention, 268\nMAC layer, 126, 127–129\nnetwork layer, 130–131\nIntrusion tolerance, 166\nIntrusion-tolerant routing in wireless sensor\nnetworks (INSENS), 480–481\nInverse discrete Fourier transform\n(IDFT), 86\nIP address security, 131, 188\nCGA, 330\nduplicated address detection, 332–334,\n337–342\ndynamic allocation, 329–332\nsecure routing protocols, 174\nspoofing attacks, 129, 388\nSUCV addresses, 330\nIPSec, 102, 175, 178\nIPSec tunnel, 281\nIPv4, 38–39, 330\nIPv6, 38–39, 330–331\nIV (nonce) management, 372, 418\nJ\nJamming, 115, 388, 445–446, See also Radio\ninterference\ndefenses against, 438, 446\nlink layer, 115, 116\nWLAN networks, 391\nJava, 160\nJunk packet forwarding, 122\nK\nKASUMI, 455\nKerberos, 203, 282, 299\nKey agreement protocol, Bluetooth,\n358–361\nencryption key and key stream, 361\ninitialization key, 358–359\nlink key, 360\nmutual entity authentication, 359–360\nunit key, 358, 362\nKey distribution, 498, See also Key\nmanagement\nCA functionality, 299–300\nID-based cryptography, 318\nIEEE 802.11i MAC layer security standard,\n132–135\nIEEE 802.1X standard, 132–133\nmessage privacy protection, 288–289\nopen research issues, 510\npredistribution, See Key predistribution\nprivacy preserving architecture, 231, See\nalso Privacy preservation\nproactive, 498–499, 502–503\nSamsung proposal for WPANs\n(KEYDS), 65\ntree-based, 499, 503, 508–510\nKey distribution center, 458\nKEYDS, 65\nKey entropy, 389\nKey-generation keys, 288–289\nKey management, 318–319, 354, 482–483,\n498–511, See also Key distribution;\nKey predistribution; specific\napplications, methods, problems\nasymmetric cryptosystem, See\nAsymmetric key cryptography\ncertification authorities, See Certificate\nauthority\nComminus proposal (Tropos Network\nand Earthlink), 404–405\ndelayed verification of signatures,\n334–337, 340–343\ndeterministic models, 421, 458–460\nduplicated address detection, 332–334,\n337–342\ndynamic address allocation, 329–332\ndynamic rekeying schemes, 423\ngeneral classification, 498\ngroupwise models, 421–422, See also\nGroup keys\nhybrid models, 421\nidentity concepts, 327–329\nIEEE 802.11i, 400\nIEEE 802.15.4 standard, 411, 419–423\nIEEE 802.1X Intel proposal, 403–404\nkey-generation keys, 288–289\nLEAP, 421, 458–460\nlocation dependent key (LDK), 326\nnetwork leaders, 334\nnon-interactive agreement and\nprogression (NIKAP), See\nNon-Interactive Key Agreement and\nProgression\non-demand exchange mechanisms, 499,\nSee also specific protocols\nopen research issues, 463, 465, 510–511\npairwise sharing, See Pairwise key\nsharing\npredistribution, See Key predistribution\nprobabilistic keying models, 318, 420,\n460–463, 501, 503–506\nrelated work, 323–326\nresiliency to node capture, 500\nresource constraint tradeoffs, 465\nrevocation, 499\n" }, { "page_number": 537, "text": "526\n■\nSecurity in Wireless Mesh Networks\nSAKM, See Simple Ad hoc Key\nManagement\nsecure AODV routing, 182\nsecurity application scenarios, 496–498\nSPINS, 325\nstatic schemes, 422–423, See also Key\npredistribution\nsymmetric cryptosystem, See Symmetric\nkey cryptography\nteam key, 177\nwireless mesh LAN security, 386, 398, 399\nwireless sensor networks and, 438,\n456–463, 493, 498–511\nZigBee architecture, 370–371, 373\nKey predistribution, 325, 498–499, 500–502,\nSee also Key management\nbalanced random predistribution,\n503–505\nIEEE 802.15.4 standard, 419–423\nLDK, 326\nlocal collaboration-based group\nre-keying, 511\nmatrix threshold (MTKP), 300\npolynomial-based, 421, 462\npolynomial threshold (PTKP), 300\nprobabilistic models, 318, 460–461\nsensor networks and, 422–423\nsymmetric key establishment\napproach, 319\nunbalanced random predistribution,\n505–506\nusing state or location information,\n501–502, 506–508\nKey ring revocation, 499\nKeys, 269, See also Key distribution; Public\nkey cryptography; Symmetric key\ncryptography\nglobal key security issues, 270\nIEEE 802.11 vulnerabilities, 389–390\nIEEE 802.11i standard, 271–275\nInternet exchange (IKE), 281\nMAC layer authentication mechanisms,\n127–129\nnon-interactive agreement protocol\n(NIKAP), See Non-Interactive Key\nAgreement and Progression\nSCAN scheme for intrusion detection, 164\nself-certified key cryptosystem, 300,\n301–302\nKiyon Mesh Network, 40\nL\nLANs, See Wireless local area networks\nLaptop-class attacks, 466\nLEAP, 421, 458–460, 469\nLIBRA, 23\nLightweight Hop-by-hop Access Protocol\n(LHAP), 275–276\nLink adaptation techniques, 15\nLink failure notification, 24\nLink key generation\nBluetooth, 358, 360\nZigBee, 427–428\nLink layer, 440\nsecurity vulnerabilities, 115, 116, 446–447\nLink Quality Source Routing (LQSR), 22,\n38, 51\nLoad and Interference Balanced Routing\nAlgorithm (LIBRA), 23\nLoad balancing, 24–26\nLoad index (LI), 25\nLocal area networks (LANs), See Wireless\nlocal area networks\nLocal collaboration-based group\nre-keying, 511\nLocalized Encryption and Authentication\nProtocol (LEAP), 421, 458–460, 469\nLocation Dependent Key (LDK)\nmanagement, 326\nLocation finding system, 439\nLocation management, 52, 53\nLocation privacy\nBluetooth, 362\nwireless mesh LAN security\nrequirements, 399\nZigBee, 375\nLogical Key Hierarchy (LKH), 458, 471\nLow-cost ripple effect attack (LORA), 124\nLQSR, 22, 38, 51\nM\nMACA-BI, 16\nMAC address security, 328, 392\nspoofing attacks, 116–117, 129, 388, 448\nwireless mesh LAN security\nrequirements, 399\nMACAW, 16\nMAC Protocol Data Unit (MPDU), 128\nMalcounts, 157–158\nMalicious collusion, proposed lightweight\nprivacy preserving solution, 247–255\nManagement frame security, 390, 391,\n397–398\nMan-in-the-middle attacks, 117–118, 135,\n365–366, 388, 391, 405\nMarshalltown WiFi network, 94\nMatrix threshold key predistribution\n(MTKP), 300\n" }, { "page_number": 538, "text": "Index\n■\n527\nMatyas-Meyer-Oseas, 424\nMD5, 455\nMedical applications, 67\nMedium access control (MAC) layer, 17–20,\nSee also specific protocols, security\nissues\ndeployment issues, 34\nIEEE 802.11i standard for security,\n132–135\nIEEE 802.15.4 network vulnerabilities,\n413\nIEEE 802.20 standard (WANs), 97–98\nmulti-channel MAC, 18–19\nmulti-radio multi-channel attacks,\n122–124\nresearch issues, 99–100\nreservation-based approach, 16–17, 89\nscheduling-based protocols, 18, 54\nsecurity attacks and vulnerabilities,\n115–118\nsecurity mechanisms, 127–130\nsingle-channel contention-based\nprotocols, 16–17\nWiMAX (IEEE 802.16) standards, 81–85\nwireless mesh LANs, 70–73\nZigBee (IEEE 802.15.4) standard, 59–60\nMedium-access Coordination Function\n(MCF), 70\nMemory constraints, sensor networks, 441\nMemory exhaustion attacks, 450\nMerkle hash-tree, 475–476\nMesh backhaul, defined, 48, 384\nMesh clients, 8–9, 49, 50\nadaptive support, 29–30\nmobility, See Mobility issues\nMesh Connectivity Layer (MCL), 38–39\nMeshDynamics QoS proposal, 62–63\nMeshed Adaptive Robust Tree (MART),\n64–65\nMesh networking products, 39–40\nMesh nodes\ndefinitions, 384\nenergy efficiency, 27\nhijacking in wireless mesh LANs,\n394–395\nMesh routers, 4, 5, 8–9, 198, See also\nRouting\nadaptive support, 29–30\ncost, 30\ndeployment, See Deployment issues\nenergy constraints, 30\nflexible deployment, 10\nhome networking, 10\nhop integrity protocol, See Hop integrity\nprotocol\nintegrating multiple network\ntechnologies, 30\nmobility management, 28–29, See also\nMobility issues\nphysical vulnerabilities, 27\nproxy RADIUS chaining, 281–282\nsecurity issues, 27\nservice availability, 33–34\nMessage authentication, See Authentication\nMessage Authentication Code (MAC), 131,\n285–286\nsensor network security, 414–415\nZigBee Alliance specification (IEEE\n802.15.4), 424\nMessage insertion or modification, See Data\nintegrity\nMessage integrity code (MIC), 118\nMeter reading applications, 92\nMetric of Interference and Channel\nSwitching (MIC), 22, 23\nMetroMesh, 77\nMetropolitan area networks (MANs), 51,\n78–94, 229\napplications, 78–79, See also WiFi\nnetworks\narchitectures, 79–80\ncentralized scheduling, 80\ndeployed solutions, 90–94\ndistributed scheduling, 80, 84\nIEEE 802.16 standards, 78\nmobility management, 85–86\nreservation-based MAC approach, 89\nrouting and QoS support, 86–90\ntargeted services, 78–79\ntransmission error correction, 85\nWiMAX (IEEE 802.16) standards, 36–37,\n80–90\nMica2 motes, 454\nMIC metric, 22, 23\nMicrosoft research, 38\nMicro-TESLA (μTESLA), 467–469, 475\nMiddleware, 31, 396, 497–498\nMisbehavior detection, 172, 266, 285–286,\n327, See also Intrusion detection\nMISTY1, 455\nMisuse detection, 150–151, 156, 161, 165\nMIT Roofnet, 37\nMobile ad hoc networks (MANETs), 47,\n382–383, See also Ad hoc networks\ncooperative anomaly detection, 155–156\nmesh network security requirements\nand, 153\nrouting security issues, 178–179\nsecure routing approach, 177–179\nsecurity challenges, 147–148\n" }, { "page_number": 539, "text": "528\n■\nSecurity in Wireless Mesh Networks\nself-organized network layer security\nsolution, 131–132\nsensor network differences, 437–438\nMobile agents, 160\nMobile Intrusion Detection system\n(MobIDS), 159–160\nMobile IP (MIP), 277–278, 325\nMobile Wireless Broadband Access\n(MWBA), 94\nMobility issues, 28–29, 50, 147–148\nadaptive EAP-TLS authentication\nsolution, 282\nadaptive support for routers\nand clients, 30\nauthentication in multi-operator mesh\nnetworks, 282–283\nhand-off management, 28–29, 52–53,\n85–86, 95–96, 98, 102, 278\nintrusion detection, 153–154\nlocation management, 52, 53\nsecurity challenges, 263–264\nWMN research issues, 102\nMoorhead WiFi network, 79\nMote-class attacks, 442\nMoving boundary-based load balancing, 25\nMulticast traffic routing, 100\nMulti-channel MAC (MMAC), 18–19\nMulti-hop wireless networks, 6, 49–50, 54,\n153, See also Mobile ad hoc\nnetworks; Wireless mesh networks;\nspecific applications, layers,\nprotocols, types\nMultiple Access Collision Avoidance By\nInvitation (MACA-BI), 16\nMultiple-input multiple-output (MIMO), 16,\n99, 100\nMulti-Radio LQSR (MR-LQSR), 22–23\nMulti-radio multi-channel (MRMC) MAC,\n19, 256\nopen security issues, 140\nrouting metrics, 21–22\nrouting protocol, 22–23\nsecurity attacks and vulnerabilities,\n122–125\nterminal access points, 263\nMulti-radio node, 384\nMulti-radio unification protocol (MUP), 19\nMutual entity authentication, Bluetooth,\n359–360\nN\nNeighborhood distributed consensus\nprotocol, 163\nNeighborhood key sharing, 128–129\nNeighborhood networking, 10–11\nNetwork-based intrusion detection systems,\n149–150\nNetwork endo-parasite attack (NEPA),\n123–124\nNetwork layer, 121, 440, 465, See also\nRouting security\ncontrol plane attacks, 119–121\ndata plane attacks, 121–122\nmulti-radio multi-channel attacks,\n122–124\nresearch issues, 100\nsecurity mechanisms, 130–132\nsensor network security, 447–449,\n465–466\nNetwork model, proposed privacy\npreserving solution, 233–236\nNetwork monitoring, 53\nNetwork partitioning attacks, 121\nNetwork technology interoperability, 31\nNonce management, 372, 418\nNon-Interactive Key Agreement and\nProgression (NIKAP), 130–131,\n297–320, 298\nAOSR routing protocol, 298, 306–310\napplication scenarios, 305\nasynchronous configuration (A-NIKAP),\n304–305, 318–319\nopen issues, 318–319\nperformance evaluation, 313–317\nrekeying, 300, 303–305, 318, 319\nrelated work, 317–318\nsecurity analysis, 310–313\nself-certified key cryptosystem, 301–302\nsynchronous configuration (S-NIKAP),\n303–304, 318–319\nNon-repudiation, 126\nrouting security issues, 176\nsensor network security requirements, 441\nNormal profiles and anomaly detection,\n151–152\nNortel mesh networking solutions, 39–40,\n94\nNtru-Encrypt, 452\nO\nOFDM, 15, 80, 86, 99, 175\nOFDMA, 80, 95, 96–97\nOff-line attacks, 389\nOLSR, 71, 175\nOnion routing, 291\nOpen-source security software libraries, 390\nOpen trust model, 367\nOptimal transmission power, 27\n" }, { "page_number": 540, "text": "Index\n■\n529\nOptimized Link State Routing (OLSR),\n71, 175\nOrthogonal Frequency Division Multiple\nAccess (OFDMA), 80, 95, 96–97\nOrthogonal Frequency Division\nMultiplexing (OFDM), 15, 80, 86,\n99, 175\nP\nPacket authentication, 275–276\nPacket forwarding misbehavior detection,\nSee Misbehavior detection\nPacket leashes, 449\nPacket scheduling, contention-based\nprotocols, 18\nPacket sniffing, 149–150\nPairwise key sharing, 419–420, 498, 501,\n509, See also Key management\nAOSR protocol, 306\nBROSK, 459–460\nCombinatorial design theory (CDT), 460\ngroupwise models, 421–422, See also\nGroup keys\nIEEE 802.15.4 data authentication, 415\ninterleaved hop-by-hop\nauthentication, 479\nLEAP, 421, 459, 469\nnon-interactive key agreement and\nprogression, 298, 300, 302–306, 318\nprobabilistic models, 420, 460–463, See\nalso Probabilistic key management\nscalability issues, 269, 438, 457\nsecure data aggregation protocol, 475\nPairwise Master Key (PMK), 133, 137,\n272, 280\nPairwise Transient Key (PTK), 133\nPANA, 280–281\nPAN coordinator, 410, 412\nPartial matching attacks, 118, 139\nPartitioned host-based load balancing, 25–26\nPassive eavesdropping, See Eavesdropping\nPathraters, 156–157\nPebblenets, 325\nPenalty-based routing algorithm, 228,\n236–239, 289\ncolluded traffic analysis, 247–255\nperformance trade-off, 244–247\nPerceived quality of service, 53\nPeripheral device security, 387\nPermutation vector (PV), 128\nPersonal area networks (PANs), 10, 56–65,\n349–376\narchitecture, 57\nbeacons, 60–61\nBluetooth security, See Bluetooth\nbroadcast scheduling, 18\nheartbeats and QoS, 53, 63\nIEEE 802.15.4 standard, See ZigBee\nIEEE 802.15.5 standard, 57, 60–65\nIEEE 802.15 standard, 16, 35–36, 352, See\nalso Bluetooth\nmeshing and UWB, 58–59\nQoS challenges, 56\nrouting and QoS support, 62–65\nrouting challenges, 56\nsecurity architecture design, 353–359\nUWB physical layer technique, 16\nZigBee (IEEE 802.15.4), See ZigBee\nPhase Shift Keying (PSK), 15\nPhysical layer, 15–17, 440\nIEEE 802.20 standard (WANs), 96–97\nnetwork capacity and, 15\nresearch issues, 99\nsecurity attacks and vulnerabilities, 115,\n445–446\nWiMAX (IEEE 802.16) standards, 86\nZigBee (IEEE 802.15.4) standard, 59\nPiconet, 352\nPico Net Controllers (PNCs), 57\nPing, 149\nPing of Death, 388\nPIN security, 362–363\nPlaintext-based secure data aggregation,\n474–477\nPoint-to-multipoint (PMP) mode, 79\nPolynomial-based encryption algorithm\n(E0), 356\nPolynomial-based key pre-distribution, 300,\n421, 462\nPolynomial threshold key predistribution\n(PTKP), 300\nPortals, 35, 70, 384\nPort control mechanisms, 395\nPower consumption, 50, See also\nBattery power\nad hoc networking issues, 7–8\npower management, 27–28, 53\nresource exhaustion attacks, 447\nsecurity challenges, 265\nsecurity overhead, 27\nsensor networks, 412, 413, 441, 511\nsleep deprivation attack, 363\nPower control message spoofing, 265\nPower management, 27–28, 53, 102,\nSee also Power consumption\nPower scavenging devices, 439\nPower spectrum density (PSD), 58\nPower units, 439\nPre-computation attacks, 118, 139\n" }, { "page_number": 541, "text": "530\n■\nSecurity in Wireless Mesh Networks\nPredictive Wireless Routing Protocol\n(PWRP), 77\nPreferred channel list (PCL), 18–19\nPrivacy homomorphism, 477–478\nPrivacy preservation, 227–258, 287–291,\n292, See also Traffic confidentiality\narchitecture, 230–232\ncollusion analysis, 247–255\nefficient key distribution, 288–289\ninformation theory, 256–257, 289\nnetwork model, 233–236\nnon-traceability, 290–291\nonion routing, 291\npenalty-based routing algorithm, 228,\n236–239, 289\nperformance trade-off, 244–247\nrelated work, 255–257\nsimulation study, 239–247\ntraffic entropy, 228, 233–236\nProactive key distribution, 498–499,\n502–503\nProbabilistic key management, 318, 420,\n460–463, 501, 503\nbalanced random predistribution,\n503–505\nunbalanced random predistribution,\n505–506\nProbabilistic stripping-based load\nbalancing, 26\nProtocol Data Unit (PDU), 128\nProxy chaining, 275, 281–282\nPseudo Hadmard Transform (PHT),\n357–358\nPseudonyms, 287\nPublic-domain software vulnerabilities,\n390\nPublic key cryptography, 269, 354, 499\ninitial authentication protocol for hop\nintegrity, 204\nmessage privacy protection, 288\nprivacy preserving architecture, 231, See\nalso Privacy preservation\nSAKM, 331, 337\nsensor networks and, 421, 438, 471,\n482, 510\nTrusted Computing AODV, 286\nwireless sensor networks and, 451–454\nPublic key infrastructure (PKI), 269, 354\nPublic safety applications, 12–13, 31–32, 79\nQ\nQuadrature Amplitude Modulation\n(QAM), 15\nQuagga, 336\nQuality of service (QoS), 53–54\ncontention-based MAC protocols, 16–17\nDiffServ, 53, 88\nDistributed Resolution Protocol, 62\nheartbeats, 53, 63\nIEEE 802.20 standard (WANs), 98\nIntServ, 53\nMAC-level research issues, 100\nMeshDynamics proposal, 62–63\nSamsung proposal, 63–65\nsensor network security and, 482–483\ntrust and reputation, 285\nWiMAX Mesh Mode, 86–90\nwireless mesh PANs, 56, 62–65\nWMR, 53, 73–76\nR\nRadio interference, 14, 262–263, See also\nJamming\nMIC metric, 22, 23\npotential solutions, 263\nSIR and channel reusability, 28\nRadio transmission, 15–16\nRADIUS, 132, 271, 275, 278, 281–282\nRAM, 441\nRate Control Protocol (RCP), 101\nRC4, 455\nRC5, 455, 470\nRC6, 455\nRCP, 101\nReactive jammer, 115\nReal-Time MAC (RTMAC), 18\nReal-time networking applications\ncontention-based MAC protocols, 16\nemergency situation deployment, 32\nvideo streaming, 11, 56, 482\nReal-Time Transport Protocol (RTCP), 101\nReduced function devices (RFDs), 57\nRekeying, non-interactive key agreement\nand progression, 300, 302–305,\n318, 319\nReplay attacks, 117–118, 129, 197, 198–199,\n391, 448\nhop integrity vs., 201, 213, 218, 224\nIEEE 802.15.4 security, 415, 419\nZigBee vulnerabilities, 374\nReputation-based security mechanisms,\n284–285, 327\nRESANE, 162\nResearch testbeds, 37\nReservation-based MAC, 16–17, 89\nResource availability, 30, See also Power\nconsumption\nResource reservation protocol (RSVP), 53\n" }, { "page_number": 542, "text": "Index\n■\n531\nRijndael, 455\nRobust Secure Network (RSN), 386, 402\nRoofnet, 37\nRoute discovery, WMR protocol for wireless\nmesh LANs, 73–74\nRoute Error (RERR) message security,\n180–181, 182, 186, 189, 193,\n308–310, 333, 335, 392\nRoute failure notification (RFN), 24\nRoute recovery, WMR protocol, 75\nRoute re-established notification (RRN), 24\nRoute Reply (RREP) security, 131, 180, 182,\n183–185, 189, 190–191, 343\nRoute Request (RREQ) security, 119, 131,\n179, 181, 183–185, 189, 190, 192,\n286, 306–308, 310–311, 343, 392\nRouting, 20–23, See also specific\napplications, protocols, or systems\ncross-layer design, 100\ndeployment issues, 34\nIEEE 802.11s WLAN protocols, 70–71\nmetrics, 20–22, 51–52, 100\nmulticast traffic, 100\nmulti-radio multi-channel attacks,\n122–124\nprotocols, 22–23, 51–52\nQoS, See Quality of service\nresearch issues, 100\nWiMAX (IEEE 802.16) QoS and wireless\nmesh MANs, 86–90\nwireless mesh LAN infrastructure, 73–76\nwireless mesh PANs, 56, 62–65\nRouting loops, 23, 121, 132, 448, See also\nRoute Request (RREQ) security\nRouting security, 171–193, 268, 496–497,\nSee also Attacks and vulnerabilities\nACK spoofing, 449\nad hoc network security, 177–179\nadministrative distances, 178, 336\nanomaly detection, 131\nAODV, 179–182, 188–189, See also\nAODV\nAOSR, 310–313, See also Ad-hoc\nOn-demand Secure Routing\nARAN, 130–131, 173\nbroadcast-based route discovery\nvulnerabilities, 392\ndirected diffusion protocol, 465–466,\n470–471\nforward-secure signature schemes, 188\ngated, 176\nhash chains, 173\nIEEE 802.11s standard, 71, 175\nimpersonation attack, 148\nimport authorization, 176–178, 334–335\nintrusion detection schemes, 154–155,\nSee also Intrusion detection\nintrusion-tolerant, 480–481\nkey distribution, See Key management\nmiddleware security, 497–498\nmobile sensor networks and, 482\nmutable and non-mutable information,\n178–179, 186, 188–189\nnetwork layer attacks and vulnerabilities,\n119–121\nnetwork layer security mechanisms,\n130–132\nnetwork monitoring, 53\nonion routing, 291\nopen issues, 187–188\npenalty-based algorithm for privacy\npreservation, 228, 236–239, 289\nprivacy preserving architecture, 232, See\nalso Privacy preservation\nprotocol comparison (table), 472\nrelated work, 172–175\nrouting protocol independence, 270\nSAODV, 131, 173, 182–193, See also\nSecure AODV\nsecure routing protocol design, 175–176\nsecurity requirements, 176–177\nsensor networks and, 413, 438, 447–449,\n465–473, 496–497\nseparate infrastructure and ad hoc\nnetwork protocols, 177–178\nSRP, 130, 173–175\nteam key, 177\ntrusted routing, 286–287\nwireless mesh LAN security\nrequirements, 399\nRSA, 342, 451–454\nRSNA, 400\nRTCP, 101\nRTMAC, 18\nRushing attacks, 119\nS\nSAFER+ block cipher, 354, 355, 357, 365\nSAKM, See Simple Ad hoc Key Management\nSalt variations of hash algorithms, 333\nSamsung QoS proposal, 63–65\nSAODV, See Secure AODV\nScalability requirements, 269\nScalable OFDMA (SOFDMA), 95\nSCAN, 163–164\nScheduling-based MAC level protocols, 18,\n54, 80, 84, 100\nSEAD, 173\nSecret Authentication Key (SAK), 127–128\n" }, { "page_number": 543, "text": "532\n■\nSecurity in Wireless Mesh Networks\nSecret exchange protocol, 198, 199, 202,\n208–214\nSecret Session Key (SSK), 127–128\nSecTrace, 285–286\nSecure AODV (SAODV), 130, 131, 173,\n182–193\ndelayed verification of signatures,\n334–337, 340–343\ndestination sequence numbers, 181, 182,\n186, 187\ndigital signatures, 182, 184–186, 187,\n190–193\nhash chains, 182–184\nsecuring error messages, 186\nSimple Ad hoc Key Management, 324,\n334, 336\nvulnerabilities and open issues, 187–188\nSecure differential data aggregation\n(SDDA), 476–477\nSecure information aggregation (SIA),\n474–475\nSecure Network Encryption Protocol\n(SNEP), 470\nSecure Routing Protocol (SRP), 130,\n173–175\nSecurity issues, 26–27, 54, 102–103,\n111–115, See also Attacks and\nvulnerabilities; Intrusion detection;\nspecific applications, mechanisms,\nmethods, protocols, or systems\ncapacity and overhead challenges,\n264–265\ndefense-in-depth, 148\nhop integrity, See Hop integrity protocol\nidentity, 327–329\nIEEE 802.11i MAC layer security standard,\n132–135\nlightweight privacy preserving solution,\nSee Privacy preservation\nmobility, See Mobility issues\nnode cooperation, 265–266, See also\nSelfish nodes\nopen issues, 139–140\nPC user practices, 146–147\npower overhead, 27\npublic-domain software\nvulnerabilities, 390\nrelated work, 172–173\nrouting, See Routing security\nspecial challenges for mesh networks,\n152–154, 263–270\nwireless sensor networks and, 413–415\nSecurity manager, 277, 370\nSecurity protocols for sensor networks,\nSee SPINS\nSecurity requirements, 262, 268–270, 494\nhop integrity, 201\nMANETs and mesh networks, 153\nsecure routing, 176–177\nsecurity architecture design, 354\nsensor networks, 443, 444, 494–496\nwireless mesh LANs, 398–399\nSecurity surveillance systems, 11–12\nSEEMesh, 69\nSelf-certified key (SCK) cryptosystem, 300,\n301–302\nSelf-configuration capability, 33, 50, See\nalso Auto-configuration\nIEEE 802.11s WLAN protocols, 72\nSelf-healing capability, 444\nSelfish nodes, 54, 113–114, 266\ndata plane attacks, 121–122\nSensing units, 439\nSensor networks, See Wireless sensor\nnetworks\nSensor nodes, 439, See also Wireless sensor\nnetworks\ncommunication range, 441\nprotocol stack, 440\nresource constraints, 441, 511\nsecurity system resiliency to\ncapture, 500\nsoftware management solutions, 497\ntamper resistance, 441, 446\nSequence numbering\nAODV and SAODV protocols, 181, 182,\n186, 187\nhop integrity check protocol, 218\nSession hijacking attack, 118, 135\nSession keys\nBROSK, 459–460\nIEEE 802.11i standard, 271–275\nSHA-1, 455\nSignal-to-interference ratio (SIR), channel\nreusability and, 28\nSignal-to-noise ratio (SNR), 15\nSignatures of attacks, 151, 165\nSimple Ad hoc Key Management\n(SAKM), 324\ndelayed verification of signatures,\n334–337\nduplicated address detection, 332–334,\n337–342\nIP address generation, 330–331\nmessage fields, 331–332\npublic key encoding, 337\nsignature encoding, 337\nSimple Network Management Protocol\n(SNMP), 160\nSimultaneous Operating Piconets, 63\n" }, { "page_number": 544, "text": "Index\n■\n533\nSingle-hop wireless networks, 6–7, See also\nWireless local area networks\nSingle-radio node, 384\nSinkhole attack, 120, 393–394,\n448, 466\nSKEME, 404\nSKKE protocol communication steps,\n428–431\nSleep deprivation attacks, 363, 374\nSleep-wake cycle aware key\npre-distribution, 502, 506–508\nSlotted Seeded Channel Hopping\n(SSCH), 19\nSmart antenna, 16, 99\nSmartDust, 441\nSNEP, 470\nSniffing, 386–387\nSnooze state, 24\nSnort, 149–150\nSPINS, 325, 438, 467, 470, 494, 497\nSpoofing attacks, 116–117, 129, 388, 448\nACK messages, 449\nauthenticator spoofing, 135–137\npower control messages, 265\nSSCH, 19\nStar-based topology, 410\nState-based key distribution, 501–502,\n506–508\nState-Based Key Hop (SBKH) protocol, 288\nState transition analysis technique (STAT),\n160–161\nStream ciphers, 127\nStreaming applications, 11, 56\nStrix Access/One, 39, 77, 93–94\nSubscriber station, 79\nSUCV addresses, 330\nSupplicant-authenticator dilemma, 396\nSurveillance applications, 11–12, 305\nSwitchable channel, 20\nSybil attack, 121, 393, 448\nSymmetric key cryptography, 269, 298, 299,\n326, See also Key distribution\nkey management, See Key management\nkey-predistribution scheme, 319, See also\nKey predistribution\nnon-interactive agreement protocol, See\nNon-Interactive Key Agreement and\nProgression\nsecure sensor network routing protocols,\n469–472\nsensor network security, 438\ntamper-resistant approaches and, 327\nwireless sensor networks and, 455–456\nZigBee Alliance specification (IEEE\n802.15.4), 424–425\nT\nTampering, 27, 446\nTamper-proofing, 446\nTamper resistance, 327, 441\nTBRPF, 20\nTCP, See Transmission Control Protocol\nTcpdump, 150\nTCPF, 24\nTDMA, 81, 99, 411\nTEA, 455\nTeam key, 177\nTempe, Arizona, WiFi system, 93\nTemporal Key (TK), 133\nintegrity protocol (TKIP), 288, 386\nTerminal access points (TAPs), 263\nTESLA, 276, 467–469, 475\nTestbed deployments, 37\nThreshold cryptography, 317–319\nThreshold secret sharing, 128, 131\nTIARA, 157\nTime Division Duplex (TDD), 36, 81\nTime Division Multiple Access (TDMA), 81,\n99, 411\nTime-division multiplexing (TDP), 447\nTime-memory trade-off attack (TMTO), 118\nTime stamp, initial authentication\nprotocol, 207\nTime synchronization, 173, 174, 175, 303\nTinyECC, 454\nTinyOS, 441, 454\nTinyPK, 454\nTLS, 203\nToken-based re-authentication, 279–280\nToken time expiration scheme, 132\nTopology Broadcast based on Reverse Path\nForwarding (TBRPF), 20\nTopology discovery, WMR protocol for\nwireless mesh LANs, 73\nTraceroute, 285–286\nTraffic admission ratio, 76\nTraffic analysis, 390\nTraffic confidentiality, 229–230, 289–290,\nSee also Privacy preservation\ncollusion analysis, 247–255\ninformation theory, 256–257\nnon-traceability, 290–291\npenalty-based routing algorithm,\n236–239\nTraffic entropy, 228, 233–236\nsimulation, 240\nTraffic padding, 257\nTraining data and anomaly detection, 151\nTransmission Control Protocol (TCP),\n23–24, 100–101\n" }, { "page_number": 545, "text": "534\n■\nSecurity in Wireless Mesh Networks\nacknowledgment frame security\nissues, 390\nad hoc network routing security, 178\nTCP-Feedback (TCPF), 24\nTransmission power, 27–28\nTransmission range, sensor nodes, 441\nTransmission rate, 15\nTransparency of security\nmechanisms, 270\nTransportation systems, 79, 94\nTransport layer, 23–24, 440\ndeployment issues, 34\nEAP-TLS, 279\nresearch issues, 100–101\nsensor network vulnerabilities, 449–450\nTrapeze, 271\nTree-based key distribution, 499, 503,\n508–510\nTropos Networks solutions, 77, 90–92,\n404–405\nTrust, 128, 283–287, 292\ndefinition, 283\nmisbehavior detection, 285–286\nmulti-operator mesh networks, 282–283\nnetwork layer security mechanisms, 130\nnon-symmetrical relations, 283–284\npacket forwarding and, 284, 285–286\nreputation and, 284–285\nrules enforcement and, 284\ntrusted routing, 286–287\nZigBee open trust model, 367\nTrust center, ZigBee, 353, 371–372\nTrusted Computing AODV\n(TCAODV), 286\nTrust model for intrusion detection,\n161–162\nTunneling attacks, 187–188, 311, 394, 466,\nSee also Wormhole attacks\nU\nUCSB MeshNet, 38\nUltra Wide Band (UWB), 15–16, 58–59, 99\nUnit key, Bluetooth, 358, 362\nUtility meter reading applications, 92\nV\nVehicle-based systems, 282\nVehicle monitoring systems, 79\nVideo game consoles, 390\nVideo streaming, 11, 56, 482\nVPN, 102\nVulnerabilities, See Attacks and\nvulnerabilities; Security issues\nW\nWar-driving, 386\nWarehousing, 66\nWatchdog, 156–157, 266\nWATCHERS, 154–155\nWater meter reading, 92\nWCETT, 21–23\nWDAP, 273–274\nWeighted Cumulative Expected\nTransmission time (WCETT), 21–23\nWeighted Radio and Load Aware (WRALA)\nmetric, 71\nWEP, See Wired Equivalent Privacy\nWide area networks (WANs), 51, 94–98\nIEEE 802.16e, 94–96\nIEEE 802.20, 94, 95, 96–99\nmobility management, 95–96\nMWBA, 94\nQoS and routing support, 98\nWiFi networks, 34\nChaska WISP, 90–91\nChittagong Access/One deployment, 93\nCorpus Christi multi-use system, 91–92\nlimitations and alternative\ntechnologies, 112\nmeter reading application, 92\nMetroMesh Networks architecture, 77\nMoorhead system, 79\nNortel’s Marshalltown case study, 94\nreputation-based systems, 285\nTempe case study, 93\nTropos Networks solutions, 90–92\nWiFi Protected Access (WPA), 116\nWiMAX, 36–37, 80–90, See also IEEE\n802.16\nIEEE 802.16e (Mobile WiMAX), 86\nMAC layer, 81–85\nmobility management, 85–86\nphysical layer, 86\nQoS in Mesh Mode, 86–90\ntransmission error correction, 85\nWi-Mesh, 69\nWindmill polynomials, 356\nWired Equivalent Privacy (WEP), 134,\n385–386\njamming attacks and, 116\nkey recovery vulnerabilities, 389–390\nWireless ad hoc networks, See Mobile ad\nhoc networks\nWireless Distribution System (WDS), 69\nWireless Dual Authentication Protocol\n(WDAP), 273–274\nWireless Internet Service Provider (WISP),\n78–79, 90\n" }, { "page_number": 546, "text": "Index\n■\n535\nWireless local area networks (WLANs), See\nalso Wireless mesh LANs\nbasic security issues, 386–387\nclient mobility and, 29, See also Mobility\nissues\nenterprise applications, 51, 66–67\nhybrid architecture for AAA, 278\nIEEE 802.11i authentication model,\n272–275, See also IEEE 802.11i\nIEEE 802.11 standard, 6, 34–35, See also\nIEEE 802.11\nmesh alternatives, 5\nmesh networking products, 40\nsimulation tools, 391\nwireless mesh LANs vs., 66\nWireless mesh LANs, 65–78, 381–406\naccess point auto-configuration, 68, 69,\nSee also Auto-configuration\napplications, 66–67\napproaches against attacks, 392–393\nattacks and vulnerabilities, 387–394,\n405–406\nattacks on networks, 387–391\nattacks on protocols, 392–394\nauthentication server location, 396–397\navailable commercial systems, 77–78\nbasic security issues, 386–387\nchallenges, 68\ncomponents and definitions, 383–385\nenterprise applications, 51\nHybrid Wireless Mesh Protocol, 392, 395\nIEEE 802.11i security protocols, 385–386\nIEEE 802.11s standard, 67–73, 400–405\nMAC protocols, 70–73\nmanagement frame security, 390, 391,\n397–398\nmesh portals, 70\nnode hijacking, 394–395\nno real mutual authorization, 396\nopen issues, 405–406\nrouting and QoS support, 73–76\nsecurity requirements, 398–399\nsupplicant-authenticator dilemma, 396\nthreats from bridged networks, 383, 395\ntraditional WLANs vs., 66\nunfairness from greedy nodes, 395–396\nWEP vulnerabilities, 389–390\nWireless mesh MANs, 78–94, See also\nMetropolitan area networks\nWireless mesh networks (WMNs), 3–6, 13,\n45–48, 198, 228, 298–299\nacademic research testbeds, 37–38\ncapacity and bandwidth, 14–15\ncharacteristics, 49–50\nIEEE standard deployments, 34–37\nindustrial research, 38–39\nMAC layer protocols, 97–98\nphysical layer standard, 96–97\nresearch issues (OSI layers), 99–102\nWireless mesh networks (WMNs),\napplications, 4, 9–13, 47–48, 50–51,\nSee also specific applications\nenterprise networking, 51, 66–67\nhealth care environments, 67\nmesh networking products, 39–40\nmeter reading, 92\npublic safety, 31–32, 79\nsensor networks, 436–437\nWiFi, See WiFi networks\nWireless mesh networks (WMNs),\narchitectures, 8–9, 48–49,\n112–113, 384\nfully and partial meshed networks, 48\nintegration with other network\ntechnologies, 30–31\nload balancing, 24–26\nMAC protocols, 17–20, See also Medium\naccess control (MAC) layer\nmulti-hop connectivity, 6, 49–50\nmulti-radio multi-channel, 19, 21–23\nPANs, 57\npeer-to-peer topology, 9\nphysical layer, 15–17, See also\nPhysical layer\nproposed privacy preserving solution,\n230–232\nrouting, See Routing\ntransport layer protocols, 23–24\nwireless mesh LANs, 67–68\nWireless mesh networks (WMNs),\ndeployment, See Deployment issues\nWireless mesh networks (WMNs), IEEE 802\nstandards, See specific IEEE standards\nWireless mesh networks (WMNs), security\nissues, See Security issues\nWireless mesh PANs, 56–65, 349–376, See\nalso Personal area networks\nWireless Mesh Routing (WMR), 53, 73–76\nWireless mesh WANs, 94–98, See also Wide\narea networks\nWireless Protected Access (WPA), 385–386\nWireless sensor networks, 435–483\nad hoc network differences, 437–438\napplications, 436–437\navailability issues, 411\nbroadcast authentication, 467–469, 510\ncluster-based networks, 409–413\ncommunication architecture, 439–440\ncontinuous stream security, 482\ncryptography, 451–456\n" }, { "page_number": 547, "text": "536\n■\nSecurity in Wireless Mesh Networks\ndata aggregation, 412\ndata aggregation security, 438,\n473–478, 497\nevent-detection applications, 413\nfuture directions, 482\nIEEE 802.15.4 security services, 415–419\nIEEE 802.15.4 standard, 409–431, See also\nIEEE 802.15.4\nintrusion detection, 478–481\nintrusion-tolerant routing (INSENS),\n480–481\nkey management, 419–423, 438, 456–463,\n493, 498–511\nlink layer vulnerabilities, 446–447\nLogical Key Hierarchy (LKH), 458, 471\nμTESLA, 467–469\nnetwork and routing layer vulnerabilities,\n447–449\nnetwork topologies, 412\nphysical layer vulnerabilities, 445–446\npower consumption, 412, 413, 441, 511\nprotocol stack, 440\npublic key models, 421, 451–454, 471,\n482, 510\nQoS and security, 482–483\nresource constraints, 440–441\nrouting security issues, 413, 438,\n447–449, 465–473, 496–497\nsecurity attacks and vulnerabilities, 438,\n443–450, 465–466\nsecurity challenges and techniques,\n413–415\nsecurity evaluation, 444\nsecurity requirements, 443, 444,\n494–496\nSPINS, 325, 438, 467, 470, 497\nsymmetric key cryptography, 455–456\nthreat model, 443–444\ntransmission range, 441\ntransport layer vulnerabilities, 449–450\nZigBee Alliance security services,\n423–431\nWitness-based data aggregation\n(WDA), 477\nWMR, 53, 73–76\nWorldwide Interoperability for Microwave\nAccess (WiMAX), 36, 80–90\nWormhole attacks, 119, 140, 306, 311–313,\n394, 449, 466\ndefenses, 131, 449\nWPA (Wireless Protected Access),\n385–386\nWRALA, 71\nX\nX.509 certificate, 203\nZ\nZebra, 336\nZigBee, 59–60, 352–353, 355, 366–376\nAES algorithm, 369–370\nbattery lifetime, 351, 353\nBluetooth differences, 351, 376\nCCM * algorithm, 367, 368–369\ncryptographic primitives, 368\ngroup keying, 372\nhash function for message\nauthentication, 424\ninitialization procedure, 374\nintegrity protection, 375\nIV (nonce) management, 372\nkey hierarchy, 370–371\nkey management, 373\nlink key derivation and confirmation,\n427–428\nlocation privacy, 375\nopen issues, 376\nreplay attacks, 374\nsecurity services for sensor networks,\n423–431\nsecurity weaknesses, 372–376\nshared secret generation, 426\nSKKE protocol communication steps,\n428–431\nsymmetric key establishment approach,\n424–425\ntrust center, 353, 371–372\nZombie computers, 388\n" }, { "page_number": 548, "text": "" } ] }