A Test-Bed for Mobile Ad-hoc Networks

Transcription

1 A Test-Bed for Mobile Ad-hoc Networks How Much Can Watchdogs Really Do? Sonja Buchegger, Cedric Tissieres, Jean-Yves Le Boudec EPFL (Swiss Federal Institute of Technology Lausanne) WMCSA, December 3, /3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 1

2 Presentation Outline Problem: Detecting Misbehavior in Mobile Ad-hoc Networks Attacks on Dynamic Source Routing (DSR) Detectability of Attacks Proposed Solution: Enhanced Passive Acknowledgment Test-Bed Performance Evaluation: Some Experimental Results Related Work Conclusions 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 2

3 Mobile Ad-hoc Networks Network of devices, no infrastructure, nodes forward packets for others. Nodes cooperate to communicate. 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 3

4 But Why Cooperate? Misbehavior Pays Off Selfish: to save power Example: No or incorrect forwarding Malicious: to attack the net Example: Route deviation Faulty: (no reason) Example: Repeating packets 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 4

5 Here s the Dilemma! Tragedy of the Commons: Free ground for everyone to let sheep graze Individually: good to put many sheep Overall: too many sheep! 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 5

6 Problem Statement How can we make a system work despite misbehavior? Which types of misbehavior are actually detectable and how? 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 6

7 Background: Dynamic Source Routing (DSR) 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 7

8 DSR - Route Request Route Request(E[A]) A B E Route Request(E[A,B]) Route Request (D[A,B]) Route Request (E[A,B]) Route Request(E[A]) Route Request(E[A,C]) C D Cache: E 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 8

9 DSR - Route Reply Route Reply(A, [E,B,A]) A B Route Reply(A, [E,B,A]) C D E Route Reply(A, [E,D,C,A]) Route Reply(A, [E,D,C,A]) Cache: E 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 9

10 DSR Data A B Data(E, [A,C,D]) C Data(E, [A,C,D]) D Data(E, [A,C,D]) E Cache: E 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 10

11 Acknowledgments in DSR Explicit ACK Passive ACK Link-layer notification 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 11

12 Enhanced Passive Acknowledgment A B PACK: Overhearing of Forwarding Tampering Fabrication In addition: Packet Reception C 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 12

13 Attacks on DSR Dropping Attacks All or partial Omit Route Error Modification Attacks Forged routing packets Added nodes Last Hop External Salvage intact routes Loops Tamper with RREQ, RREP Decrease TTL Fabrication Attacks Forged RERR Spoofed RREQ Forged RREP Frequent RREQ Timing Attacks RREP disproportionally fast 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 13

14 Test-Bed Components Piconet with PACK, enhanced PACK, and attacks APE Netfilter with promiscuous mode Pcmcia-cs with promiscuous mode Setup: Laptops with Linux kernels and , Orinoco Classic Gold b cards 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 14

15 Test-Bed Architecture 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 15

16 Implemented Example Attacks Header Modification Selfish Attacks Remove from RREP RERR modification Attacks work! Malicious Attacks Change Source Route RERR destination Attacks work! Partial Dropping Attack works! RERR Fabrication Attack works! 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 16

17 Experimental Results 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 17

18 Experimental Results II 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 18

19 Related Work: Economic Incentives Forwarding is rewarded. Target: selfish/rational nodes Examples: nuglets/counters, Crowcroft, Sprite Solution only for the non-forwarding type of misbehavior. 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 19

20 Related Work: Secure Routing Using Cryptography to secure route discovery Target: malicious nodes Examples: Ariadne, SRP, S-AODV, BISS Solution only for route discovery. Nodes can still deviate traffic or drop packets. 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 20

21 Related Work: Reputation Systems 1 In MANET or P2P: Keep track of misbehaving nodes, exclude them Target: misbehaving nodes regardless of reason Examples: Watchdog, CORE, Context, OCEAN, ID, Aberer, SECURE Either Use only first-hand information, so only detect neighbors, or are vulnerable to spurious ratings, or assume trust transitivity, or only consider negative (positive) information 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 21

22 Related Work: Reputation Systems 2 E-Commerce History of transactions for future choice of partners Target: human decision makers, agents Examples: E-Bay Centralized 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 22

23 Solution Proposal: CONFIDANT Target both routing and forwarding misbehavior Regardless whether selfish, faulty, or malicious Be able to detect misbehavior before meeting (use second-hand information) Cope with spurious ratings Fully distributed 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 23

24 Purpose of CONFIDANT CONFIDANT detects misbehaving nodes by means of observation or reports about several types of attacks and thus allows nodes to route around misbehaved nodes and to isolate misbehaved nodes from the network, so that misbehavior does not pay off, cannot continue, and routes are functional. 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 24

25 Misbehavior F B C A D E G 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 25

26 Publication F B C A D E G 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 26

27 Isolation and Rerouting F B C A D E G 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 27

28 Conclusions Watchdogs can work well Enhanced Passive ACK can detect quite a lot Watchdogs with enhanced PACK can give useful input to misbehavior detection and reputation systems Need to do larger test-bed experiments to find limitations, false positives Make code and documentation freely available 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 28

29 Watch This Space! Code and Documentation will be available from Soon. 12/3/04 A Test-Bed for Mobile Ad-hoc Networks S. Buchegger, C. Tissieres, JY. Le Boudec 29