Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Dependent Link Padding Algorithms for Low Latency Anonymity SystemsW. Wang, M. Motani, V. SrinivasanCCS 2008Presented by B. Choics6461 at MTUMotivation●Low latency anonymity systems are vulnerable to traffic analysis attacks●One way to thwart such an attack is to use dummy traffic●Understanding of the cost and effectiveness is low●Where to start?Things to think ...●Scope: entire network, tier-1 AS, tier-2 AS, tier-3 AS, ...–Tarzan?●Effectiveness of dummy traffic–Linkability from a suspect input to any suspect output to be:●Minimized?●Randomized●Equalized?●Cost: genuine traffic vs. dummy trafficBackground●Independent link padding–Scope: one hop–Output pattern: pre-determined regardless of input–Straightforward output patterns: constant, exponential (Poisson)●Dependent link padding–Scope: one hop–Output pattern: determined online depending on input–How to produce output with given input?Intuition●Independent link padding:–Very strong resistance against traffic analysis–Low bandwidth utilization●Dependent link padding–Maybe strong enough to resist traffic analysis–Flexible bandwidth utilization–Can there be a good framework on DLP?Assumptions●Input flows are about of the same rate in Poisson●All packets belong to a flow (link) are sent to the same output flow (link)●Single anonymity server (mix) with a strict delay bound ●The mix does not drop any packet●All output links show the same output to maximize the anonymityMixMatching packetsProposed DLP algorithmExample of outputClaims●The dummy traffic is minimized (max efficiency)●Sending rate proportional to log(m)–M: the number of input flows●Multi-hop: upper-bounded delay x hops●Experiment on the sending rateExperiment on delay boundComparison with ILPsReal Traffic (2003)Packet drop ratesDrawback of DLPDrawback of
View Full Document