1. Toward Prevention of Traffic Analysis Fengfeng Tu 11/26/01

2. Discussion Outline What is traffic analysis? What is traffic analysis? What are traffic analysis attacks? What are traffic analysis attacks? How to prevent traffic analysis attacks? How to prevent traffic analysis attacks? Problems Problems Conclusion Conclusion

3. Traffic Analysis Monitor the network traffic Monitor the network traffic  e.g. log files, webpage hits, etc. http://www.openwebscope.com/samples/math_yal e_stats.html http://www.openwebscope.com/samples/math_yal e_stats.html http://www.openwebscope.com/samples/math_yal e_stats.html http://www.openwebscope.com/samples/math_yal e_stats.html Gain useful information from statistical analysis Gain useful information from statistical analysis  Who communicates with whom, when, how long, where?  Who is interested in what contents?

4. Traffic Analysis Attacks An adversary is doing traffic analysis An adversary is doing traffic analysis  e.g., earlier versions of SSH protocol Communication Pattern Communication Pattern  Sender-recipient matchings  Traffic volume, traffic shape  Duration Examples of sensitive info Examples of sensitive info  Possible corporate takeover  Importance of communicating parties

5. Anti-Traffic Analysis Anonymizer Anonymizer AT&T Crowds AT&T Crowds Onion Routing Onion Routing  Pentagon hides behind onion wraps Freedom Freedom Most are Chaum Mix-like Most are Chaum Mix-like

6. Chaum Mixes David Chaum. “Untraceable Electronic Mail, Return Addresses, and Digital Pseudonyms”, Communication of the ACM, 1981. David Chaum. “Untraceable Electronic Mail, Return Addresses, and Digital Pseudonyms”, Communication of the ACM, 1981. Mix nodes are intermediate processors that a message goes through. Mix nodes are intermediate processors that a message goes through. Purpose - hide the correspondences between the incoming and outgoing messages. Purpose - hide the correspondences between the incoming and outgoing messages.

7. How it works? The message will be sent through a series of mix nodes: 1, 2, …,d-1, d. The user encrypts the message with node d’s private key, then encrypts the result with (d-1)’s private key and so on. The message will be sent through a series of mix nodes: 1, 2, …,d-1, d. The user encrypts the message with node d’s private key, then encrypts the result with (d-1)’s private key and so on. MIX 1MIX 2 K 2 (R 2, K Y (R 0, M), A Y ), A 2 X K 1 (R 1, K 2 (R 2, K Y (R 0, M), A Y ), A 2 ), A 1 K Y (R 0, M), A Y Y

8. How it works? (Cont’) The mix nodes receive a certain number of these messages which they decrypt, randomly reorder and send to the next node The mix nodes receive a certain number of these messages which they decrypt, randomly reorder and send to the next node The order of outgoing messages changed, so it is nearly impossible to correlate a message that comes in with a message that goes out. The order of outgoing messages changed, so it is nearly impossible to correlate a message that comes in with a message that goes out.

9. A Mix Node

10. How it works? (Cont’) Link-to-link encryption is not sufficient. Link-to-link encryption is not sufficient.  Mix nodes are not trusted (insider attacks). Why do we need random numbers? Why do we need random numbers? MIX 1 X Y K 1 (K Y (M), A Y ), A 1 K Y (M), A Y Encrypt it with K 1 => K 1 (K Y (M), A Y ) = ?

11. Characteristics Sender/Recipient Anonymity - each mix node only knows the previous and next node in a received message’s route. Sender/Recipient Anonymity - each mix node only knows the previous and next node in a received message’s route. Constant message length Constant message length  Large message are chopped into short ones with a specific constant length  Padding if the message is too small Each message is processed by a Mix only once Each message is processed by a Mix only once

12. A Simple Example

13. Problems? Brute Force Attacks Brute Force Attacks Duration of a communication can be observed. Duration of a communication can be observed. An extreme case: An extreme case:

14. Dummy Traffic All users send messages at all times All users send messages at all times  All users start and end their communication at the same time  Long communication is chopped into slices If a user has nothing to send, it sends random numbers indistinguishable from real (encrypted) messages. If a user has nothing to send, it sends random numbers indistinguishable from real (encrypted) messages. Reduce delay Reduce delay

15. Problems Imposing rigid structure on user communications Imposing rigid structure on user communications Dummy messages waste resources Dummy messages waste resources Delays at the Mixes. Delays at the Mixes. Cost of nested encryption Cost of nested encryption

16. Routing Issues Rerouting or Multi-path routing to improve network utilization Rerouting or Multi-path routing to improve network utilization  Reduce the dummy traffic volume 5Mbps Dummy traffic 2.5Mbps All are real traffic

17. Rerouting Host-based rerouting Host-based rerouting i. Compute the shortest path for each flow ii. Select a flow randomly or according to a sequence defined in advance. iii. Remove the traffic requirement for that flow iv. Reroute flow to reduce an objective function value, with routing paths for all other flows fixed v. Go to step (ii) until all flows have been examined at least once, but no further improvements are possible

19. Problems Solving a system of linear inequalities Solving a system of linear inequalities  Linear programming The computation is centralized to avoid local hot spot problem The computation is centralized to avoid local hot spot problem  Too expensive: consider all flows  Vulnerable to single-point failure

20. Conclusion Anonymity and Unobservability are hard to achieve in Internet Anonymity and Unobservability are hard to achieve in Internet The situation is worse in wireless (ad hoc) networks The situation is worse in wireless (ad hoc) networks  The media is open  Link transmission interference  Multi-path routing needed  Distributed algorithm

21. Literature Research David Chaum. Untraceable Electronic Mail, Return Addresses, and Digital Pseudonyms. Communication of the ACM, 1981. David Chaum. Untraceable Electronic Mail, Return Addresses, and Digital Pseudonyms. Communication of the ACM, 1981. J. Raymond. Traffic Analysis: Protocols, Attacks, Design Issues, and Open Problems. J. Raymond. Traffic Analysis: Protocols, Attacks, Design Issues, and Open Problems. O. Berthold, et al. Project “Anonymity and Unobservability in the Internet”. CFP 2000. O. Berthold, et al. Project “Anonymity and Unobservability in the Internet”. CFP 2000. M. Reed, et al. Anonymous Connections and Onion Routing. IEEE Journal on Special Areas in Communications, May 1998 M. Reed, et al. Anonymous Connections and Onion Routing. IEEE Journal on Special Areas in Communications, May 1998 R. Newman, et al. High Level Prevention of Traffic Analysis. 7 th Annual Computer Security and Applications Conference, Dec. 1991. R. Newman, et al. High Level Prevention of Traffic Analysis. 7 th Annual Computer Security and Applications Conference, Dec. 1991. S. Jiang, et al. Routing in Packet Radio Networks to Prevent Traffic Analysis. Proc. of IEEE Information Assurance and Security Workshop, West Point, NY, June 2000. S. Jiang, et al. Routing in Packet Radio Networks to Prevent Traffic Analysis. Proc. of IEEE Information Assurance and Security Workshop, West Point, NY, June 2000. http://netcamo.cs.tamu.edu/ http://netcamo.cs.tamu.edu/ http://netcamo.cs.tamu.edu/