1. 1 Security and Misbehavior Handling in Wireless Ad Hoc Networks Nitin H. Vaidya University of Illinois at Urbana-Champaign nhv@uiuc.edu http://www.crhc.uiuc.edu/~nhv © 2005 Nitin Vaidya

2. 2 Notes  Coverage not exhaustive. Only a few example schemes discussed  Only selected features of various schemes are typically discussed. Not possible to cover all details in this tutorial  Some protocol specs have changed over time, and the slides may not reflect the most current specifications  Jargon used to discuss a scheme may occasionally differ from that used in the original papers  Names in brackets, as in [Xyz00], refer to a document in the list of references  Abbreviation MAC used to mean either Medium Access Control or Message Authentication Code – implied meaning should be clear from context

3. 3 Outline  Introduction to ad hoc networks  Selected routing and MAC protocols  Key management in wireless ad hoc networks  Secure communication in ad hoc networks  Misbehavior at the MAC layer  Misbehavior at the network layer  Anomaly detection

4. 4 Mobile Ad Hoc Networks (MANET)

5. 5 Mobile Ad Hoc Networks  Formed by wireless hosts which may be mobile  Without (necessarily) using a pre-existing infrastructure  Routes between nodes may potentially contain multiple hops

6. 6 Mobile Ad Hoc Networks  May need to traverse multiple links to reach a destination A B C D

7. 7 Mobile Ad Hoc Networks (MANET)  Mobility causes route changes A B C D

8. 8 Why Ad Hoc Networks ?  Ease of deployment  Speed of deployment  Decreased dependence on infrastructure

9. 9 Many Applications  Personal area networking  cell phone, laptop, ear phone, wrist watch  Military environments  soldiers, tanks, planes  Civilian environments  taxi cab network  meeting rooms  sports stadiums  boats, small aircraft  Emergency operations  search-and-rescue  policing and fire fighting

10. 10 Many Variations  Fully Symmetric Environment  all nodes have identical capabilities and responsibilities  Asymmetric Capabilities  transmission ranges and radios may differ  battery life at different nodes may differ  processing capacity may be different at different nodes  speed of movement  Asymmetric Responsibilities  only some nodes may route packets  some nodes may act as leaders of nearby nodes (e.g., cluster head)

11. 11 Many Variations  Traffic characteristics may differ in different ad hoc networks  bit rate  timeliness constraints  reliability requirements  unicast / multicast / geocast  host-based addressing / content-based addressing / capability-based addressing  May co-exist (and co-operate) with an infrastructure- based network

12. 12 Many Variations  Mobility patterns may be different  people sitting at an airport lounge  New York taxi cabs  kids playing  military movements  personal area network  Mobility characteristics  speed  predictability direction of movement pattern of movement  uniformity (or lack thereof) of mobility characteristics among different nodes

13. 13 Challenges  Limited wireless transmission range  Broadcast nature of the wireless medium  Hidden terminal problem (see next slide)  Packet losses due to transmission errors  Mobility-induced route changes  Mobility-induced packet losses  Battery constraints  Potentially frequent network partitions  Ease of snooping on wireless transmissions (security hazard)

14. 14 Hidden Terminal Problem BCA Nodes A and C cannot hear each other Transmissions by nodes A and C can collide at node B Nodes A and C are hidden from each other

15. 15 Research on Mobile Ad Hoc Networks Variations in capabilities & responsibilities X Variations in traffic characteristics, mobility models, etc. X Performance criteria (e.g., throughput, energy, security) = Significant research activity

16. 16 The Holy Grail  A one-size-fits-all solution  Perhaps using an adaptive/hybrid approach that can adapt to situation at hand  Difficult problem  Many solutions proposed trying to address a sub-space of the problem domain

17. 17 Outline  Introduction to ad hoc networks  Selected routing and MAC protocols  Key management in wireless ad hoc networks  Secure communication in ad hoc networks  Misbehavior at the MAC layer  Misbehavior at the network layer  Anomaly detection

18. 18 Unicast Routing in Mobile Ad Hoc Networks

19. 19 Why is Routing in MANET different ?  Host mobility  link failure/repair due to mobility may have different characteristics than those due to other causes  Rate of link failure/repair may be high when nodes move fast  New performance criteria may be used  route stability despite mobility  energy consumption

20. 20 Unicast Routing Protocols  Many protocols have been proposed  Some have been invented specifically for MANET  Others are adapted from previously proposed protocols for wired networks  No single protocol works well in all environments  some attempts made to develop adaptive protocols

21. 21 Routing Protocols  Proactive protocols  Determine routes independent of traffic pattern  Traditional link-state and distance-vector routing protocols are proactive  Reactive protocols  Maintain routes only if needed  Hybrid protocols

22. 22 Trade-Off  Latency of route discovery  Proactive protocols may have lower latency since routes are maintained at all times  Reactive protocols may have higher latency because a route from X to Y may be found only when X attempts to send to Y  Overhead of route discovery/maintenance  Reactive protocols may have lower overhead since routes are determined only if needed  Proactive protocols can (but not necessarily) result in higher overhead due to continuous route updating  Which approach achieves a better trade-off depends on the traffic and mobility patterns

23. 23 Reactive Routing Protocols

24. 24 Routing Protocols  Proactive protocols for ad hoc networks are often derived from link state or distance vector routing protocols  But with some optimizations  We will not discuss proactive protocols in detail  Before discussing an example reactive protocol, let us consider “flooding” as a routing protocol

25. 25 Flooding for Data Delivery  Sender S broadcasts data packet P to all its neighbors  Each node receiving P forwards P to its neighbors  Sequence numbers used to avoid the possibility of forwarding the same packet more than once  Packet P reaches destination D provided that D is reachable from sender S  Node D does not forward the packet

26. 26 Flooding for Data Delivery B A S E F H J D C G I K Represents that connected nodes are within each other’s transmission range Z Y Represents a node that has received packet P M N L

27. 27 Flooding for Data Delivery B A S E F H J D C G I K Represents transmission of packet P Represents a node that receives packet P for the first time Z Y Broadcast transmission M N L

28. 28 Flooding for Data Delivery B A S E F H J D C G I K Node H receives packet P from two neighbors: potential for collision Z Y M N L

29. 29 Flooding for Data Delivery B A S E F H J D C G I K Node C receives packet P from G and H, but does not forward it again, because node C has already forwarded packet P once Z Y M N L

30. 30 Flooding for Data Delivery B A S E F H J D C G I K Z Y M Nodes J and K both broadcast packet P to node D Since nodes J and K are hidden from each other, their transmissions may collide  Packet P may not be delivered to node D at all, despite the use of flooding N L

31. 31 Flooding for Data Delivery B A S E F H J D C G I K Z Y Node D does not forward packet P, because node D is the intended destination of packet P M N L

32. 32 Flooding for Data Delivery B A S E F H J D C G I K Flooding completed Nodes unreachable from S do not receive packet P (e.g., node Z) Nodes for which all paths from S go through the destination D also do not receive packet P (example: node N) Z Y M N L

33. 33 Flooding for Data Delivery B A S E F H J D C G I K Flooding may deliver packets to too many nodes (in the worst case, all nodes reachable from sender may receive the packet) Z Y M N L

34. 34 Flooding for Data Delivery: Advantages  Simplicity  May be more efficient than other protocols when rate of information transmission is low enough that the overhead of explicit route discovery/maintenance incurred by other protocols is relatively higher  this scenario may occur, for instance, when nodes transmit small data packets relatively infrequently, and many topology changes occur between consecutive packet transmissions  Potentially higher reliability of data delivery  Because packets may be delivered to the destination on multiple paths

35. 35 Flooding for Data Delivery: Disadvantages  Potentially, very high overhead  Data packets may be delivered to too many nodes who do not need to receive them  Potentially lower reliability of data delivery  Flooding uses broadcasting -- hard to implement reliable broadcast delivery without significantly increasing overhead –Broadcasting in IEEE 802.11 MAC is unreliable  In our example, nodes J and K may transmit to node D simultaneously, resulting in loss of the packet –in this case, destination would not receive the packet at all

36. 36 Flooding of Control Packets  Many protocols perform (potentially limited) flooding of control packets, instead of data packets  The control packets are used to discover routes  Discovered routes are subsequently used to send data packet(s)  Overhead of control packet flooding is amortized over data packets transmitted between consecutive control packet floods  Several protocols based on this (Examples: DSR, AODV)

37. 37 Dynamic Source Routing (DSR) [Johnson96]  When node S wants to send a packet to node D, but does not know a route to D, node S initiates a route discovery  Source node S floods Route Request (RREQ)  Each node appends own identifier when forwarding RREQ

38. 38 Route Discovery in DSR B A S E F H J D C G I K Z Y Represents a node that has received RREQ for D from S M N L

39. 39 Route Discovery in DSR B A S E F H J D C G I K Represents transmission of RREQ Z Y Broadcast transmission M N L [S] [X,Y] Represents list of identifiers appended to RREQ

40. 40 Route Discovery in DSR B A S E F H J D C G I K Node H receives packet RREQ from two neighbors: potential for collision Z Y M N L [S,E] [S,C]

41. 41 Route Discovery in DSR B A S E F H J D C G I K Node C receives RREQ from G and H, but does not forward it again, because node C has already forwarded RREQ once Z Y M N L [S,C,G] [S,E,F]

42. 42 Route Discovery in DSR B A S E F H J D C G I K Z Y M Nodes J and K both broadcast RREQ to node D Since nodes J and K are hidden from each other, their transmissions may collide N L [S,C,G,K] [S,E,F,J]

43. 43 Route Discovery in DSR B A S E F H J D C G I K Z Y Node D does not forward RREQ, because node D is the intended target of the route discovery M N L [S,E,F,J,M]

44. 44 Route Discovery in DSR  Destination D on receiving the first RREQ, sends a Route Reply (RREP)  RREP is sent on a route obtained by reversing the route appended to received RREQ  RREP includes the route from S to D on which RREQ was received by node D

45. 45 Route Reply in DSR B A S E F H J D C G I K Z Y M N L RREP [S,E,F,J,D] Represents RREP control message

46. 46 Route Reply in DSR  Route Reply can be sent by reversing the route in Route Request (RREQ) only if links are guaranteed to be bi-directional  To ensure this, RREQ should be forwarded only if it received on a link that is known to be bi-directional  If unidirectional (asymmetric) links are allowed, then RREP may need a route discovery for S from node D  Unless node D already knows a route to node S  If a route discovery is initiated by D for a route to S, then the Route Reply is piggybacked on the Route Request from D.  If IEEE 802.11 MAC is used to send data, then links have to be bi-directional (since Ack is used)

47. 47 Dynamic Source Routing (DSR)  Node S on receiving RREP, caches the route included in the RREP  When node S sends a data packet to D, the entire route is included in the packet header  hence the name source routing  Intermediate nodes use the source route included in a packet to determine to whom a packet should be forwarded

48. 48 Data Delivery in DSR B A S E F H J D C G I K Z Y M N L DATA [S,E,F,J,D] Packet header size grows with route length

49. 49 When to Perform a Route Discovery  When node S wants to send data to node D, but does not know a valid route node D

50. 50 Route Error (RERR) B A S E F H J D C G I K Z Y M N L RERR [J-D] J sends a route error to S along route J-F-E-S when its attempt to forward the data packet S (with route SEFJD) on J-D fails Nodes hearing RERR update their route cache to remove link J-D

51. 51 Unicast Routing Protocols  We will use DSR as the example routing protocol in much of our discussion

52. 52 Outline  Introduction to ad hoc networks  Selected routing and MAC protocols  Key management in wireless ad hoc networks  Secure communication in ad hoc networks  Misbehavior at the MAC layer  Misbehavior at the network layer  Anomaly detection

53. 53 Medium Access Control Protocols

54. 54 Medium Access Control  Wireless channel is a shared medium  Need access control mechanism to avoid interference  MAC protocol design has been an active area of research for many years [Chandra00]

55. 55 MAC: A Simple Classification Wireless MAC CentralizedDistributed Guaranteed or controlled access Random access IEEE 802.11

56. 56 ABC Hidden Terminal Problem  Node B can communicate with A and C both  A and C cannot hear each other  When A transmits to B, C cannot detect the transmission using the carrier sense mechanism  If C transmits, collision will occur at node B

57. 57 MACA Solution for Hidden Terminal Problem [Karn90]  When node A wants to send a packet to node B, node A first sends a Request-to-Send (RTS) to A  On receiving RTS, node A responds by sending Clear-to-Send (CTS), provided node A is able to receive the packet  When a node (such as C) overhears a CTS, it keeps quiet for the duration of the transfer  Transfer duration is included in RTS and CTS both ABC

58. 58 Reliability  Wireless links are prone to errors. High packet loss rate detrimental to transport-layer performance.  Mechanisms needed to reduce packet loss rate experienced by upper layers

59. 59 A Simple Solution to Improve Reliability  When node B receives a data packet from node A, node B sends an Acknowledgement (Ack). This approach adopted in many protocols [Bharghavan94,IEEE 802.11]  If node A fails to receive an Ack, it will retransmit the packet ABC

60. 60 IEEE 802.11 Wireless MAC  Distributed and centralized MAC components  Distributed Coordination Function (DCF)  Point Coordination Function (PCF)  DCF suitable for multi-hop ad hoc networking  DCF is a Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA) protocol

61. 61 IEEE 802.11 DCF  Uses RTS-CTS exchange to avoid hidden terminal problem  Any node overhearing a CTS cannot transmit for the duration of the transfer  Uses ACK to achieve reliability  Any node receiving the RTS cannot transmit for the duration of the transfer  To prevent collision with ACK when it arrives at the sender  When B is sending data to C, node A will keep quite ABC

62. 62 Collision Avoidance  CSMA/CA: Wireless MAC protocols often use collision avoidance techniques, in conjunction with a (physical or virtual) carrier sense mechanism  Carrier sense: When a node wishes to transmit a packet, it first waits until the channel is idle.  Collision avoidance: Nodes hearing RTS/CTS stay silent for specified duration. Once channel becomes idle, the node waits for a randomly chosen duration before attempting to transmit.

63. 63 CFABED RTS RTS = Request-to-Send IEEE 802.11 Pretending a circular range

64. 64 CFABED RTS RTS = Request-to-Send IEEE 802.11 NAV = 10 NAV = remaining duration to keep quiet

65. 65 CFABED CTS CTS = Clear-to-Send IEEE 802.11

66. 66 CFABED CTS CTS = Clear-to-Send IEEE 802.11 NAV = 8

67. 67 CFABED DATA DATA packet follows CTS. Successful data reception acknowledged using ACK. IEEE 802.11

68. 68 IEEE 802.11 CFABED ACK

69. 69 CFABED ACK IEEE 802.11 Reserved area (not necessarily circular in practice)

70. 70 Backoff Interval  Backoff intervals used to reduce collision probability  When transmitting a packet, choose a backoff interval in the range [0,cw]  cw is contention window  Count down the backoff interval when medium is idle  Count-down is suspended if medium becomes busy  When backoff interval reaches 0, transmit RTS

71. 71 IEEE 802.11 DCF Example data wait B1 = 5 B2 = 15 B1 = 25 B2 = 20 data wait B1 and B2 are backoff intervals at nodes 1 and 2 cw = 31 B2 = 10

72. 72 Backoff Interval  The time spent counting down backoff intervals is a part of MAC overhead  Choosing a large cw leads to large backoff intervals and can result in larger overhead  Choosing a small cw leads to a larger number of collisions (when two nodes count down to 0 simultaneously)

73. 73  Since the number of nodes attempting to transmit simultaneously may change with time, some mechanism to manage contention is needed  IEEE 802.11 DCF: contention window cw is chosen dynamically depending on collision occurrence

74. 74 Binary Exponential Backoff in DCF  When a node fails to receive CTS in response to its RTS, it increases the contention window  cw is doubled (up to an upper bound)  When a node successfully completes a data transfer, it restores cw to Cw min  cw follows a sawtooth curve

75. 75 Security and Misbehavior

76. 76 Issues  Hosts may be misbehave or try to compromise security at all layers of the protocol stack

77. 77 Transport Layer (End-to-End Communication)  How to secure end-to-end communication?  Need to know keys to be used for secure communication  May want to anonymize the communication

78. 78 Network Layer Misbehaving hosts may create many hazards  May disrupt route discovery and maintenance: Force use of poor routes (e.g., long routes)  Delay, drop, corrupt, misroute packets  May degrade performance by making good routes look bad

79. 79 MAC Layer  Disobey protocol specifications for selfish gains  Denial-of-service attacks

80. 80 Scope of this Tutorial  Overview of selected issues at various protocol layers  Not an exhaustive survey of all relevant problems or solutions

81. 81 Outline  Introduction to ad hoc networks  Selected routing and MAC protocols  Key management in wireless ad hoc networks  Secure communication in ad hoc networks  Misbehavior at the MAC layer  Misbehavior at the network layer  Anomaly detection

82. 82 Key Management

83. 83 Key Management  In “pure” ad hoc networks, access to infrastructure cannot be assumed  Network may also become partitioned  In “hybrid” networks, however, if access to infrastructure is typically available, traditional solutions can be extended with relative ease

84. 84 Certification Authority  Certification Authority (CA) has a public/private key pair, with public key known to all  CA signs certificate binding public keys to other nodes  A single CA may not be enough – unavailability of the CA (due to partitioning, failure or compromise) will make it difficult for nodes to obtain public keys of other hosts  A compromised CA may sign erroneous certificates

85. 85 Distributed Certification Authority [Zhou99]  Use threshold cryptography to implement CA functionality jointly at n nodes. The n CA servers collectively have a public/private key pair  Each CA only knows a part of the private key  Can tolerate t compromised servers  Threshold cryptography: (n,t+1) threshold cryptography scheme allows n parties to share the ability to perform a cryptographic operation (e.g., creating a digital signature)  Any (t+1) parties can perform the operation jointly  No t or fewer parties can perform the operation

86. 86 Distributed Certification Authority [Zhou99]  Each server knows public key of other servers, so that the servers can communicate with each other securely  To sign a certificate, each server generates a partial signature for the certificate, and submits to a combiner  To protect against a compromised combiner, use t+1 combiners

87. 87 Self-Organized Public Key Management [Capkun03]  Does not rely on availability of CA  Nodes form a “Certificate Graph”  each vertex represents a public key  an edge from K u to K w exists if there is a certificate signed by the private key of node u that binds K w to the identity of some node w. KuKu KwKw (w,K w ) Pr Ku

88. 88 Self-Organized Public Key Management [Capkun03]  Four steps of the management scheme  Step 1: Each node creates its own private/public keys. Each node acts independently

89. 89 Self-Organized Public Key Management  Step 2: When a node u believes that key K w belongs to node w, node u issues a public-key certificate in which K w is bound to w by the signature of u  u may believe this because u and w may have talked on a dedicated channel previously  Each node also issues a self-signed certificate for its own key  Step 3: Nodes periodically exchange certificates with other nodes they encounter  Mobility allows faster dissemination of certificates through the network

90. 90 Self-Organized Public Key Management  Step 4: Each node forms a certificate graph using the certificates known to that node Authentication: When a node u wants to verify the authenticity of the public key K v of node v, u tries to find a directed graph from K u to K v in the certificate graph. If such a path is found, the key is authentic.

91. 91 Self-Organized Public Key Management  Misbehaving hosts may issue incorrect certificates  If there are mismatching certificates, indicates presence of a misbehaving host (unless one of the mismatching certificate has expired)  Mismatching certificates may bind same public key for two different nodes, or same node to two different keys  To resolve the mismatch, a “confidence” level may be calculated for each certificate chain that verifies each of the mismatching certificates  Choose the certificate that can be verified with high confidence – else ignore both certificates

92. 92 TESLA Broadcast Authentication [Perrig]  How to verify authenticity of broadcast packets?  Use Message Authentication Code (MAC) for each message, using a shared secret key  But with broadcast, all receivers need to know the shared key, and any of them can then impersonate the sender  Use digital signature with asymmetric cryptography  Computationally expensive  Use asymmetric cryptography to bootstrap symmetric cryptography solution  TESLA

93. 93 TESLA  Uses one-way hash chains: Starting with initial value s 0, use one-way function F to general a sequence of values s 1 = F(s 0 ), s 2 = F(s 1 ), …, s n = F(s n-1 ).  Knowing an earlier value in the chain, a latter value can be determined, but not vice-versa  Use the values in reverse order, starting from s n-1  Order of use opposite the order of generation  Distribute s n to all nodes with verifiable authenticity  Use digital signature (this is the “bootstrap” step)  Nodes need to know the source’s public key

94. 94 TESLA  Messages sent during period i include Message Authentication Code (MAC) computed using another one-way function of s i  The key s i is revealed after a key disclosure delay of d intervals  On receiving a message in interval i, a node X waits for d-1 additional intervals for the key to be revealed)  When s i is revealed, node X can verify that s i+1 = F(s i ) to determine authenticity of s i

95. 95 TESLA  Authenticity of s i can be determined so long as node X knows some s k with k>i  Allows for loss of revealed keys during broadcast operation  Once a key is revealed, anyone can try to impersonate the sender using that key  To avoid this, TESLA assumes loose time synchronization  Each receiver can place an upper bound on the sender’s clock  The error needs to be small compared to key disclosure delay

96. 96 TESLA  If impersonator I receives key s i from source S first, and sends a packet to R impersonating S, R will find the packet valid only if  The packet timestamp is smaller than the upper bound R places on the time at S, and  Now, the upper bound when S sends key s i will be at least i+d (since the key is not released until interval i+d)  So if R only accepts packets sent with timestamp i but received when the upper bound on S’s clock < i+d, there is no way an impersonator can pass above conditions (provided clock error small compared to d) S R I

97. 97 TESLA  Advantage: Use of asymmetric cryptography required only initially (to distribute initial key using signatures) Further communication uses MAC  Disadvantage: Messages can only be authenticated after delay d

98. 98 Outline  Introduction to ad hoc networks  Selected routing and MAC protocols  Key management in wireless ad hoc networks  Secure communication in ad hoc networks  Misbehavior at the MAC layer  Misbehavior at the network layer  Anomaly detection

99. 99 Secure Communication

100. 100 Secure Communication  With the previously discussed mechanisms for key distribution, it is possible to authenticate the assignment of a public key to a node  This key can then be used for secure communication  The public key can be used to set up a symmetric key between a given node pair as well  TESLA provides a mechanism for broadcast authentication when a single source must broadcast packets to multiple receivers

101. 101 Secure Communication  Sometimes security requirement may include anonymity  Availability of an authentic key is not enough to prevent traffic analysis  We may want to hide the source or the destination of a packet, or simply the amount of traffic between a given pair of nodes

102. 102 Traffic Analysis  Traditional approaches for anonymous communication, for instance, based on MIX nodes or dummy traffic insertion, can be used in wireless ad hoc networks as well  However, it is possible to develop new approaches considering the broadcast nature of the wireless channel

103. 103 Mix Nodes [Chaum]  Mix nodes can reorder packets from different flows, insert dummy packets, or delay packets, to reduce correlation between packets in and packets out M1BM2E A M3C D G F

104. 104 Mix Nodes  Node A wants to send message M to node G. Node A chooses 2 Mix nodes (in general n mix nodes), say, M1 and M2 M1BM2E A M3C D G F

105. 105 Mix Nodes  Node A transmits to M1 message K1(R1, K2(R2, M)) where Ki() denotes encryption using public key Ki of Mix i, and Ri is a random number M1BM2E A M3C D G F

106. 106 Mix Nodes  M1 recovers K2(R2,M) and send to M2 M1BM2E A M3C D G F

107. 107 Mix Nodes  M2 recovers M and sends to G M1BM2E A M3C D G F

108. 108 Mix Nodes  If M is encrypted by a secret key, no one other than G or A can know M  Since M1 and M2 “mix” traffic, observers cannot determine the source-destination pair without compromising M1 and M2 both

109. 109 Alternative Mix Nodes  Suppose A uses M2 and M3 (not M1 and M2)  Need to take fewer hops  Choice of mix nodes affects overhead M1BM2E A M3C D G F

110. 110 Mix Node Selection  Intelligent selection of mix nodes can reduce overhead [Jiang04]  With mobility, the choice of mix nodes may have to be modified to reduce cost  However, change of mix selection has the potential for divulging more information

111. 111 Traffic Mode Detection  Consider a node pair A and D. Depending on the “mode” of operation, the traffic rate from A to D is either R1 or R2.  To avoid detection of the mode, node A may always send at rate max (R1, R2) inserting dummy traffic if necessary [Venkatraman93]  This is an end-to-end approach, since it can be implemented entirely at source & destination of a flow

112. 112 Traffic Mode Detection  Now consider two flow A-D and E-F  Mode 1: A-D rate R1E-F rate R2 Mode 2: A-D rate R2 E-F rate R1  End-to-end cover: A-D and E-F both at rate max (R1,R2)  Link BC carries traffic 2*max (R1,R2) ABCD E F Max(R1,R2) 2 * Max(R1,R2)

113. 113 Traffic Mode Detection  If we can encrypt link layer traffic in ad hoc networks, then a “link” cover mode can be used, such that each link carries fixed traffic independent of traffic mode  Reduces resource usage ABCD E F Max(R1,R2) on each link except BC R1+ R2 on link BC

114. 114 Traffic Mode Detection  Insertion of dummy traffic on a per-link basis “cheaper” than end-to-end [Radosavljevic92,Jiang01]  But need to take into account rates of different flows to determine suitable level of padding  Also, need link layer encryption to disallow differentiation of different flows at the link layer

115. 115 Traffic Mode Detection  Mode 1: A-D rate R1E-F rate R2 Mode 2: A-D rate R2 E-F rate R1  Need Max(R1,R2) on all links, since the two flows do not share links  Node B transmits 2 * Max(R1,R2) traffic ABD E F

116. 116 Traffic Mode Detection  Node-level dummy packet insertion cheaper, if we can hide link-level receiver of the packets  Without the dummy traffic, node B forwards traffic R1+R2 independent of the mode  Node-level insertion: Maintain rates Max(R1,R2) at nodes A and E, and rate R1+R2 at node B ABD E F

117. 117 Traffic Mode Detection  Node B needs to be able to remove dummy packets  Recipient of traffic from node B needs to be hidden  Additional mechanisms can be designed for this [Jiang05]

118. 118 Outline  Introduction to ad hoc networks  Selected routing and MAC protocols  Key management in wireless ad hoc networks  Secure communication in ad hoc networks  Misbehavior at the MAC layer  Misbehavior at the network layer  Anomaly detection

119. 119 Misbehavior at the MAC Layer

120. 120 MAC Layer Misbehavior Wireless channel Access Point AB Nodes are required to follow Medium Access Control (MAC) rules Misbehaving nodes may violate MAC rules Wireless channel Access Point CD

121. 121 Example  We will illustrate MAC layer misbehavior with example misbehaviors that can occur with IEEE 802.11 DCF protocol  For ease of discussion, we sometimes refer to nodes communicating with an “access point”, but the discussion applies equally to nodes transmitting to any node in an ad hoc network acting as their receiver

122. 122 Some Possible Misbehaviors  Causing collisions with other hosts’ RTS or CTS [Raya]  Those hosts will exponentially backoff on packet loss, giving free channel to the misbehaving host

123. 123 Possible Misbehaviors: “Impatient” Transmitters  Smaller backoff intervals [Kyasanur]  Shorter Interframe Spacings [Raya]

124. 124 “Impatient” Transmitters  Backoff from biased distribution  Example: Always select a small backoff value Transmit wait B1 = 1 B2 = 20 Transmit wait B2 = 19 B1 = 1 Misbehaving node Well-behaved node

125. 125 Impatient Transmitters  We will discuss the case of hosts that choose “too small” backoff intervals  But other cases of hosts waiting too little before talking can be handled analogously

126. 126 Goals [Kyasanur03]  Diagnose node misbehavior  Catch misbehaving nodes  Discourage misbehavior  Punish misbehaving nodes

127. 127 Potential Approaches  Watch idle times on the channel to detect when hosts wait too little  Design protocols that improve the ability to detect misbehavior  Protocols that discourage misbehavior [Konorski] Certain game-theoretic approaches

128. 128 Passive Observation [Kyasanur03] (Conceptually Simplest Solution)  802.11 dictates that each host must be idle for a certain duration between transmissions  The duration can be expressed as (K + v) where K is a constant, and v is chosen probabilistically from a certain distribution  K due to inter-frame spacing  v due to randomly chosen backoff intervals

129. 129 Passive Observation  The observer can measure the idle time on the channel and determine whether the idle time is drawn from the above distribution  If the observed idle time is smaller than expected, then misbehavior can be detected [Kyasanur03]  [Cagalj05] presents an implementation based on this approach

130. 130 Passive Observation  With this approach, a receiver can try to diagnose behavior of nodes trying to send packets to the receiver Wireless channel Access Point A

131. 131 Issues  Wireless channel introduces uncertainties  Not all hosts see channel idle at the same time  AP1 sees channel busy, but A sees it as idle Wireless channel AP 1 A Wireless channel AP 2 B

132. 132 Issues  Spatial channel variations bound the efficacy of misbehavior detection mechanisms  Many existing proposals ignore channel variation when performing evaluations, making the evaluations less reliable

133. 133 Issues  Receiver does not know exact backoff value chosen by sender  Sender chooses random backoff  Hard to distinguish between maliciously chosen small values and a legitimate value

134. 134 Potential Solution: Use long-term statistics [Kyasanur]  Observe backoffs chosen by sender over multiple packets  Selecting right observation interval difficult

135. 135 An Alternative Approach  Remove the non-determinism

136. 136 An Alternative Approach  Receiver provides backoff values to sender  Receiver specifies backoff for next packet in ACK for current packet  Modification does not significantly change 802.11 behavior  Backoffs of different nodes still independent Uncertainty of sender’s backoff eliminated

137. 137 Modifications to 802.11 R provides backoff B to S in ACK B selected from [0,CW min ] DATA Sender S Receiver R CTS ACK(B) RTS S uses B for backoff RTS B

138. 138 Protocol steps Step 1: For each transmission:  Detect deviations: Decide if sender backed off for less than required number of slots  Penalize deviations: Penalty is added, if the sender appears to have deviated Goal: Identify and penalize suspected misbehavior  Reacting to individual transmission makes it harder for the cheater to adapt to the protocol

139. 139 Protocol steps Step 2: Based on last W transmissions:  Diagnose misbehavior: Identify misbehaving nodes Goal: Identify misbehaving nodes with high probability  Reduce impact of channel uncertainties  Filter out misbehaving nodes from well-behaved nodes

140. 140 Detecting deviations  Receiver counts number of idle slots B obsr Condition for detecting deviations: B obsr <  B (0 <  <= 1) Sender S Receiver R ACK(B) RTS Backoff B obsr

141. 141 Penalizing Misbehavior When B obsr <  B, penalty P added  P proportional to  B– B obsr ACK(B+P) CTS DATA Total backoff assigned = B + P B obsr Sender S Receiver R ACK(B) RTS Actual backoff < B

142. 142 Penalty Scheme issues  Misbehaving sender has two options  Ignore assigned penalty  Easier to detect  Follow assigned penalty  No throughput gain  With penalty, sender has to misbehave more for same throughput gain

143. 143 Diagnosing Misbehavior  Total deviation for last W packets used  Deviation per packet is B – B obsr  If total deviation > THRESH then sender is designated as misbehaving  Higher layers / administrator can be informed of misbehavior

144. 144 Summary of Performance Results  Persistent misbehavior detected with high accuracy Accuracy increases with misbehavior  Accuracy depends on channel conditions  Accuracy not 100% due to channel variations

145. 145 Variations – Multiple Observers  In an ad hoc networks, a node can only diagnose, on its own, misbehavior by senders in its vicinity  Potential for error due to channel variations  Different hosts can cooperate to improve accuracy  Open problem: How to cooperate? How to “merge” information to arrive at a diagnosis?

146. 146 Other Approaches  Game theory  Incentive-based mechanisms

147. 147 MAC Selfishness: Game-Theoretic Approach  [MacKenzie] addresses selfish misbehavior in Aloha networks  Nodes can choose arbitrary access probabilities  Assign cost c for a transmission attempt Utility of a successful transmission = 1-c Utility of an unsuccessful transmission = -c Utility of no attempt = 0  MacKenzie’s contribution is to show that there exists a Nash equilibrium strategy

148. 148 MAC: Selfishness  Others have also attempted game-theoretic solutions [Konorski,Cagalj05]  Limitation: Game-theoretic solutions (so far) assume that all hosts see identical channel state  Not realistic  Limits usefulness of solutions

149. 149  Use payment schemes, charging per packet  Misbehaving hosts can get more throughput, but at a higher cost This solution does not ensure fairness Also, misbehaving node can achieve lower delay at no extra cost This suggests that per-packet payment is not enough Need to factor delay as well (harder) Incentive-Based Mechanisms [Zhong02]

150. 150 Outline  Introduction to ad hoc networks  Selected routing and MAC protocols  Key management in wireless ad hoc networks  Secure communication in ad hoc networks  Misbehavior at the MAC layer  Misbehavior at the network layer  Anomaly detection

151. 151 Network Layer Misbehavior

152. 152 Network Layer Misbehavior  Many potential misbehaviors have been identified in various papers  We will discuss selected misbehaviors, and plausible solutions

153. 153 Drop/Corrupt/Misroute  A node “agrees” to join a route (for instance, by forwarding route request in DSR) but fails to forward packets correctly  A node may do so to conserve energy, or to launch a denial-of-service attack, due to failure of some sort, or because of overload

154. 154 Watchdog Approach [Marti]  Verify whether a node has forwarded a packet or not B DC E A B sends packet to C

155. 155 Watchdog Approach [Marti]  Verify whether a node has forwarded a packet or not  B can learn whether C has forwarded packet or not  B can also know whether packet is tampered with if no per-link encryption B DC E A C forwards packet to D B overhears C Forwarding the packet

156. 156 Watchdog Approach: Buffering & Failure Detection  Forwarding by C may not be immediate: B must buffer packets for some time, and compare them with overheard packets Buffered packet can be removed on a match  If packet stays in buffer at B too long, a “failure tally” for node C is incremented  If the failure rate is above a threshold, C is determined as misbehaving, and source node informed

157. 157 Impact of Collisions  If A transmits while C is forwarding to D, A will not know  Failure tally at C is not reliable. Include a margin for such errors (which may be exploited by misbehaving hosts) B DC E A C forwards packet to D

158. 158 Reliability of Reception Not Known  Even if B sees the transmission from C, it cannot always tell whether D received the packet reliably  Misbehaving C may reduce power such that B can receive from C, but D does not (provided path loss to D is higher) B DC E A C forwards packet to D

159. 159 Channel Variations May Cause False Detection  If channel quality between B and C changes often, B may not overhear packets forwarded by C  This will increase C’s failure tally at B  May cause false misbehavior accusation B DC E A

160. 160 Malicious Reporting  Host D may be a good node, but C may report that D is misbehaving  Source cannot tell whether this report is accurate  If the destination sends acknowledgement to source for the received packets, and if the forward-reverse routes are disjoint, this misbehavior (by C) may be caught

161. 161 Collusion  If C forwards packets to D, but fails to report when D does not forward packets, the source node cannot determine who is misbehaving B DC E A Collusion hard to detect in many other schemes as well

162. 162 Misdirection of Packets  C forwards packets, but to the wrong node!  With DSR, B knows the next hop after C, so this misbehavior may be detected  With other hop-by-hop forwarding protocols, B cannot detect this B DC E A F

163. 163 Directional Transmissions  Directional transmissions make it difficult to use Watchdog  Power control for improved capacity or energy efficiency can create difficulties as well B DC E A B cannot hear C’s transmission to D

164. 164 Watchdog + Pathrater [Marti]  “Pathrater” is run by each node. Each node assigns a rating to each known node  Previously unknown nodes assigned “neutral” rating of 0.5  Rating assigned to nodes suspected of misbehaving are set to large negative value  Other nodes have positive ratings (between 0 and 0.8)  Ratings of well-behaved nodes increase over time up to a maximum  So a temporary misbehavior can be overcome by sustained good behavior  Routes with larger cumulative node ratings preferred

165. 165 Watchdog: Summary  Can detect misbehaving hosts, although not always; false detection possible as well  Misbehaving hosts not punished  Effectively rewarded, by not sending any more traffic through them  Potential modification: Punishment could be to not forward any traffic from the misbehaving hosts

166. 166 Hosts Bearing Grudges: CONFIDANT Protocol [Buchegger]  Motivated by “The Selfish Gene” by Dawkins (1976)  Consider three types of birds  “Suckers” – Birds that always groom parasites off other birds’ heads  “Cheats” – Birds that never help other birds  “Grudgers” – Birds that do not help known cheaters  If bird population starts out with only suckers and cheats, both categories become extinct over time  If bird population contains grudgers, eventually they dominate the population, and others become extinct

167. 167 Hosts Bearing Grudges  Applying the “grudgers” concept to ad hoc networks  Each node determines whether its neighbor is misbehaving Similar to the previous scheme  A node ALARMs its “friends” when a misbehaving hosts is detected  Each node maintains reputation ratings for other nodes that are reduced on receipt of ALARMs  Ratings improve with time – a cheater can rehabilitate itself

168. 168 Hosts Bearing Grudges: Issues  How to decide on friends?  What if “friends” cheat?

169. 169 Hosts Bearing Grudges: Summary  Reputation-based scheme  Nodes prefer to route through & for nodes with higher reputation  Interesting concept, but cannot circumvent the difficulties in diagnosing misbehavior accurately

170. 170 Exploiting Path Redundancy [Xue04]  Design routing algorithms that can deliver data despite misbehaving nodes  “Tolerate” misbehavior by using disjoint routes  Prefer routes that deliver packets at a higher “delivery ratio”

171. 171 Exploiting Path Redundancy  Alternate routes: AFGE, ABCDE, ABFGE, ABCGE B D G E A F C

172. 172 Exploiting Path Redundancy  Misbehaving host F drops packets  Delivery ratio poor on routes AFGE, ABFGE, better on ABCDE, ABCGE B D G E A F C

173. 173 Best-Effort Fault Tolerant Routing (BFTR) – Modified DSR [Xue04]  The target of a route discovery is required to send multiple route replies (RREP)  The source can discover multiple routes (all are deemed feasible initially) (1) The source chooses a feasible route based on the “shortest path” metric (2) The source uses this route until its delivery ratio falls below a threshold (making the route infeasible) (3) If existing route is deemed infeasible, go to (1)

174. 174 BFTR: Issues  A route may look infeasible due to temporary overload on that route  The source may settle on a poorer (but feasible) route  No direct mechanism to differentiate misbehavior from lower capacity routes  This is both an advantage, and a potential shortcoming

175. 175 Information Dispersal [Rabin89]  Map the N bit information F to n pieces, each N/m in size, such that any m pieces suffice to reconstruct original information Total size = n/m * N  Divide information F into N/m sequences of length m S1 = (b 1, …, b m ) S2 = (b m+1, …, b 2m ) …

176. 176 Information Dispersal  Choose n vectors a i = (a i1, …, a im ) Such that any set of m different vectors are linearly independent  Let Fi = (c i1, c i2, …, c iN/m ) 1<= i <= n where c ik = a i. S k Example: c i1 = a i.b 1 + a i2.b 2 + … + a im. b m

177. 177 Information Dispersal [Rabin89]  Given m pieces, say, F 1, …, F m, we can reconstruct F as follows  Let A = (a ij ) 1<=i,j<= m  A. S k ’ = (c 11, c 21, …, c m1 )’ ’ denotes transpose Thus, knowing A and F i = (c i1, c i2, …, c iN/m ), we can recover S

178. 178 Information Dispersal to Tolerate Misbehavior [Papadimitratos03]  Choose n node-disjoint paths to send the n pieces of information  Use a route rating scheme (based on delivery ratios) to select the routes  Acknowledgements for received pieces are sent  The missing pieces retransmitted on other routes  Need to be able to detect whether packets are tampered with

179. 179 Route Tampering Attack  A node may make a route appear too long or too short by tampering with RREQ in DSR  By making a route appear too long, the node may avoid the route from being used  This would happen if the destination replies to multiple RREQ in DSR  By making a route appear too short, the node may make the source use that route, and then drop data packets (denial of service)

180. 180 Node Insertion B A S E F H J D C G I K Z Y M N L [S,E,P,Q,F] [S,E]

181. 181 Node Deletion B A S E F H J D C G I K Z Y M N L [S,G,K] [S,C,G]

182. 182 Route Tampering Attack  Useful to allow detection of route tampering  Solution: Protect route accumulated in RREQ from tampering Removal or insertion of nodes should both be detected

183. 183 Ariadne [Hu]: Detecting Route Tampering  Source-Destination S-D pairs share secret keys Ksd and Kds for each direction of communication  One-way hash function H available  MAC = Message Authentication Code (MAC) computed using MAC keys

184. 184 Ariadne [Hu]: Detecting Route Tampering  Let RREQ’ denote the RREQ that would have been sent in unmodified DSR  Source S broadcasts RREQ = RREQ’,h 0,[] where h0 = HMAC Ksd (RREQ’)  When a node X receives an RREQ = (RREQ’, h i, [m list]) it broadcasts RREQ, m i+1 where RREQ = (RREQ’, h i+1, [m list]), m i+1 where h i+1 = H(X, h i ) and m i+1 =HMAC Kx (RREQ)

185. 185 Ariadne  If D receives an RREQ that came via route S, A, B, C, then D should have received h = H(C, H(B, H(A, HMAC Ksd (initial RREQ’))))  Knowing H and Ksd, and the node identifiers appended in the RREQ, D can verify accuracy of received h  Relies on the inability to invert function H  A mismatch indicates tampering with h or node list  A match indicates that the h value corresponds to the node-list Not enough to know whether the node-list is accurate  If no tampering detected in h, send RREP including node-list and m-list, and HMAC for this information

186. 186 Ariadne  Node D sends the RREP to node C (first node on reverse route)  Node C forwards to the next node towards the source, but also appends its key Kc to the message  One key used per route discovery (TESLA mechanism). S can verify authenticity of this key  Alternate mechanisms: Use pair-wise shared secret keys, or signatures using authentic public keys  Node S receives all the keys, and also the m-list in RREP  S can verify that all m values in the m-list are accurate, in addition to the HMAC computed by D  If all check out, then no tampering, else discard RREP

187. 187 Ariadne  If HMAC checks, then no one tampered with the node-list and m-list in the RREP  If m-list checks, then the m values were computed by legitimate nodes when RREQ forwarded  If all OK, accept RREP  Use of m-list ensures that a host cannot tamper with the RREP  Route in RREP is the route taken by RREQ and RREP

188. 188 Ariadne: Issues  Ensuring that RREQ and RREP follow the known route does not ensure that the nodes on the route will deliver packets correctly  So this is not a sufficient solution (and some might argue, not necessary!)

189. 189 Wormhole Attack [Hu]  In this attack, the attacker makes a wireless “link” appear in the network when there isn’t one  The attacker may achieve this by using an out-of- band channel, or a channel that cannot be detected by other hosts  Not necessarily detrimental, since the additional link can improve performance  But the attacker may cause the network to funnel traffic through this link, giving the attacker control on the fate of the traffic

190. 190 Wormhole Attack [Hu]  Host X can forward packets from F and E unaltered  Hosts F and E will seem “adjacent” to each other B D X E A F C

191. 191 Wormhole Attack [Hu]  With DSR, RREQ via AFXE will likely arrive at E soonest  The RREQ will contain route AFE  When RREP from E reaches A, it will start using AFE  The fact that AFE really is AFXE will not be detected B D X E A F C

192. 192 Wormhole Attack [Hu]  With DSR, RREQ via AFXE will likely arrive at E soonest  The RREQ will contain route AFE  When RREP from E reaches A, it will start using AFE  The fact that AFE really is AFXE will not be detected B D X E A F C

193. 193 Wormhole Attack [Hu]  Subsequently when A sends data along AFE, node X will not forward the data to E B D X E A F C

194. 194 Wormhole Attack: Issues  Not that simple to launch an undetected wormhole attack  If node F can “see” someone else sending packets with F specified as sender, the attack is detected  Transmissions from X must be invisible to F B D X E A F C

195. 195 Wormhole Attack: Issues  Transmissions from X must be invisible to F  Use directional transmissions at X to forward packets  Difficult for X to guarantee that F will not see its transmissions (depends on beamforms, multipath) B D X E A F C

196. 196 Wormhole Attack: Issues  Transmissions from X must be invisible to F  Out-of-band collusion between two attackers X and Y  Difficult for Y to guarantee that F will not see its transmissions B D X E A F C Y

197. 197 Wormhole Attack: Issues  Timing: F may expect an “immediate ACK”  In the absence of authentication, X can ACK packets to F without having delivered them to E  With authentication, this is difficult B D X E A F C

198. 198 Timing Issue  Alternatively, the attacker must be able to forward bits as soon as it starts receiving them from F  X transmits to E while receiving from F on the same channel  If no delays introduced, E and F may not detect the attack B D X E A F C

199. 199 Detected Attack If timing issue cannot be resolved by the attacker ….  If X cannot deliver a timely ACK, the link E  F will appear broken to E (because no ACK when expected)  Thus, even though E appears to receive RREQ from F, it cannot deliver packets to F  The attack will make the link F-E seem unidirectional (unreliable broadcast from F to E works, but not reliable unicast from E to F).  Mechanisms to handle unidirectional links (“blacklist”) can potentially suffice

200. 200 Other Detection Mechanisms: Geographical Leashes  Geographical Leashes: Each transmission from a host should be allowed to propagate over a limited distance  If E and F are too far, F should reject packets that seem to be transmitted by E, even if received reliably  Need an estimate of distance between E and F (GPS locations + mobility during packet transmission)

201. 201 Geographical Leashes [Hu]  Difficulty: Packets may travel along non line-of-sight paths  Hard to predict the actual “distance” traveled by the transmissions  Difficulty: A related problem is that physically close hosts may not be able to communicate directly (because of obstacles)  The attacker may still introduce a tunnel (wormhole) between these hosts  However, the attacker needs the information that the two hosts cannot see each other – difficult to get this information

202. 202 Temporal Leashes  Assume tight clock synchronization (e.g., GPS)  Sender timestamps the packet, and receiver determines the delay since the packet was sent  If delay too large, reject the packet  The timestamps must be protected by some authentication mechanism or signature

203. 203 Wormhole Attack: Summary  Not clear that this attack is easy to launch undetected The attacker needs knowledge of propagation to be sure of avoiding detection  Solutions dealing with unidirectional links may suffice in some cases

204. 204 Outline  Introduction to ad hoc networks  Selected routing and MAC protocols  Key management in wireless ad hoc networks  Secure communication in ad hoc networks  Misbehavior at the MAC layer  Misbehavior at the network layer  Anomaly detection

205. 205 Anomaly Detection

206. 206 Anomaly Detection  Anomaly detection: Detect deviation from “normal” behavior  Need to characterize “normal”  Normal behavior hard to characterize accurately  Need to be able to determine when observed behavior departs significantly from the norm  Avoid false positives  The MAC layer approach for detecting deviation from “normal” distribution of contention window parameters can be considered an “anomaly detection” scheme

207. 207 Anomaly Detection in Ad Hoc Networks [Zhang00]  Anomaly detection may also be useful at other layers, particularly, network layer  How to characterize “normal” routing protocol behavior?  Some of the routing mechanisms we discussed earlier do detect specific forms of abnormal behavior, but a more generic approach is desired  Can we design a protocol-independent anomaly detection mechanism? Not clear

208. 208 Anomaly Detection  We limit our discussion here  Wireless harder than wired networks due to spatial and temporal variations

209. 209 Conclusions

210. 210 Conclusion  Security an important consideration for widespread deployment of wireless ad hoc networks  We discussed a sampling of topics in security and misbehavior in ad hoc networks  Some issues are similar to those in wired networks  The differences from wired network arise due to  Shared nature of the wireless channel with variations over space/time  Inability to rely on access to “infrastructure”  Ease of intrusion (relative to wired networks)

211. 211 Conclusion  A lot of interesting research ongoing  One concern is that not all attacks are equally likely  Attackers will typically go after the weakest feature  Nevertheless an important area of research with potential for future applications

212. 212 Some Relevant Conferences/Workshops  ACM Wireless Security Workshop (WiSe) – held at ACM MobiCom last few years  Traditional security conferences (Security and Privacy, DSN, etc.)  Networking conferences: ACM MobiCom, ACM MobiHoc, IEEE INFOCOM, etc.

213. 213 Thanks! www.crhc.uiuc.edu/wireless nhv@uiuc.edu

214. 214 References  [Bharghavan94] MACAW: A Media Access Protocol for Wireless LANs, Vaduvur Bharghavan, Alan Demers, Scott Shenker, Lixia Zhang, SIGCOMM, 1994  [Buchegger] S. Buchegger and J. Le Boudec, Nodes Bearing Grudges: Towards Routing, Security, Fairness, and Robustness in Mobile Ad Hoc Networks,' in Proceedings of the Tenth Euromicro Workshop on Parallel, Distributed and Network-based Processing, IEEE Computer Society, January 2002.  [Cagalj05] M. Cagalj, S. Ganeriwal, I. Aad, and J. P. Hubaux : On Selfish Behavior in CSMA/CA Ad Hoc Networks, to appear at Infocom 20  [Capkun93] S. Capkun, L. Buttyan, and J. P. Hubaux, "Self-Organized Public- Key Management for Mobile Ad Hoc Networks“ IEEE Transactions on Mobile Computing, Vol. 2, Nr. 1 (January - March 2003)  [Chandra00] A. Chandra, V. Gummalla, and J. O. Limb, "Wireless Medium Access Control Protocols," IEEE Commun. Surveys [online], available at: http://www.comsoc.org/pubs/surveys, 2nd Quarter 2000.  [Chaum] D. Chaum, Untraceable Electronic Mail, Return Addresses, and Digital Pseudonyms", Communications of the ACM, 1981.  [IEEE 802.11] IEEE 802.11 Specification, IEEE

215. 215 References  [Hu02] Y. Hu, A. Perrig, and D. Johnson, ``Ariadne: A secure on-demand routing protocol for ad hoc networks,'' in The 8th ACM International Conference on Mobile Computing and Networking, MobiCom 2002, pp.~12--23, September 2002.  [Hu03] Y.-C. Hu, A. Perrig, and D. B. Johnson, ``Packet leashes: A defense against wormhole attacks in wireless networks,'' in Proceedings of IEEE INFOCOM'03, (San Francisco, CA), April 2003.  [Jiang04] S. Jiang, N. H. Vaidya and W. Zhao, A Mix Route Algorithm for Mix- Net in Wireless Ad Hoc Networks, IEEE International Conference on Mobile Ad- hoc and Sensor Systems (MASS), October 2004.  [Jiang01] S. Jiang, N. H. Vaidya, W. Zhao, Preventing traffic analysis in packet radio networks, DISCEX 2001.  [Jiang05] S. Jiang, N. H. Vaidya, W. Zhao, in preparation, 2005  [Johnson] David B. Johnson and David A. Maltz. Protocols for Adaptive Wireless and Mobile Networking, IEEE Personal Communications, 3(1):34-42, February 1996.  [Karn90] MACA - A New Channel Access Method for Packet Radio. Appeared in the proceedings of the 9th ARRL Computer Networking Conference, London, Ontario, Canada, 1990  [Konorski] J. Konorski, Multiple access in ad-hoc wireless LANs with noncooperative stations, NETWORKING 2002

216. 216 References  [Kyasanur], Pradeep Kyasanur and N. H. Vaidya, Selfish MAC Layer Misbehavior in Wireless Networks, to appear in the IEEE Transactions on Mobile Computing.  [Kyasanur03] P. Kyasanur and N. H. Vaidya, Detection and Handling of MAC Layer Misbehavior in Wireless Networks, Dependable Computing and Communications Symposium (DCC) at the International Conference on Dependable Systems and Networks (DSN), June 2003.  [Papadimitratos03] Papadimitratos and Haas, Secure message transmission in mobile ad hoc networks, Ad Hoc Networks journal, 2003.  [Perrig] A. Perrig, TESLA Project, http://www.ece.cmu.edu/~adrian/tesla.html.  [Rabin89] M. O. Rabin, Efficient dispersal of information for security, load balancing, and fault tolerance, J. ACM 38, 335-348 (1989)  [Marti00] S. Marti, T. J. Giuli, K. Lai, and M. Baker, ``Mitigating routing misbehavior in mobile ad hoc networks,'' in ACM International Conference on Mobile Computing and Networking (MobiCom), pp. 255--265, 2000.  [Radosavljevic92] B. Radosavljevic, B. Hajek, Hiding traffic flow in communication networks, MILCOM 1992.

217. 217 References  [Raya] M. Raya, J.-P. Hubaux, and I. Aad, `DOMINO: A System to Detect Greedy Behavior in IEEE 802.11 Hotspots.,'' in Proceedings of ACM MobiSys, Boston - MA, 2004  [Venkatraman93] B. R. Venkatraman and N. E. Newman-Wolfe, Transmission schedules to prevent traffic analysis, Ninth Annual Computer Security and Applications Conferences, 1993.  [Xue04] Yuan Xue and Klara Nahrstedt, "Providing Fault-Tolerant Ad-hoc Routing Service in Adversarial Environments," in Wireless Personal Communications, Special Issue on Security for Next Generation Communications, Kluwer Academic Publishers, vol 29, no 3-4, pp 367-388, 2004  [Zhong02] Sprite: A Simple, Cheat-Proof, Credit-Based System for Mobile Ad- Hoc Networks, Infocom 2003  [Zhou99] Securing Ad Hoc Networks, Lidong Zhou, Zygmunt J. Haas, IEEE Network, 1999