Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12GAP – Practical anonymous networkingKrista Bennett, Christian GrothoffPET’03Presented by B. Choi in cs6461Computer ScienceMichigan TechIntroduction●Applications of anonymous communication–Electronic payment, voting, auction, email, and web browsing–One still not mentioned: File Sharing!●What’s the differences?–Query and reply–Widely spread P2P application already●Which platform?–GNUnetP2P file sharing?●File sharing, what is it?–You share I share–Equal rights and equal responsibilities–Napster in 2000●P2P in general–Structured–Unstructured–Hybrid●Security and trust are the primary concernGNUnet (my impression)●Unstructured system–How to join the system●Well known node (distributed registry)●Obtain partial membership from the registry●Leave? - when you want!–Query forwarding●Random selection of next nodes●Multiple forwarding at each forward●Time-to-live to remove loops–Reply●Encoded blocks●Content migrationQuery and reply●Fundamental difference from other applications–More query the higher chances to hit a copy–One-to-many (file sharing) vs. initiator-and-responder (other applications)–Potentially many replies with different blocks of the target file (movie or music titles)●What needs to be anonymous?–Who (identity) is looking for which file?–Who (identity) is responding to which query?–Sender and receiver anonymity?Basic decisions●To have or not? Cover traffic–Not chosen as in many other P2P-based anonymity systems (Crowds, MorphMix, Tarzan?)–Why not?●Churn●Content migration●Probabilistic responding●Dynamism●Adversary model–External passive, internal active (colluding nodes)Main idea●GAP: GNUnet anonymity protocols (my guess)●For a given time window, a node–Creates n queries–Forwards m foreign queries–Indirect k foreign queries out of m–Anonymity of a node is n/(n + m – k)–Has to maintain a routing status for each indirect-ed foreign query (how long?)–Forwarding/indirecting to a random selection of nodes–Decision based on local situations (workload)Indirecting (not new) vs. forwardingThoughts on indirect/forward●What anonymity is affected?–Originator?–Responder?–Forwarder?●Why would one choose forward?–Better efficiency of the node●What about the system in general?–Better efficiency?–Higher vulnerability (vs. easy content migration?)Thoughts on hops-to-live●Traditional hops-to-live would leak much information●Solution: TTL --> time window to process a reply–TTL + the local time–Another hole for the adversary?–Delay to process exceptionally soon reply?Measuring anonymityDiscussions●GAP: individual node chooses whether to exchange portions of its own anonymity for its own efficiency without impacting the security of other nodes●Statefulness!●Any other application not studied in light of anonymity?–Social networks (Facebook, etc,..)–Skype!–Instant
View Full Document
Unlocking...