applications for measurement improving anonymity online
play

Applications for Measurement: Improving Anonymity Online Rishab - PowerPoint PPT Presentation

Applications for Measurement: Improving Anonymity Online Rishab Nithyanand | Rachee Singh | Shinyoung Cho | Phillipa Gill Stony Brook University 1 Anonymity on the Internet Tor Network 2 Anonymity on the Internet Does not know the source


  1. Applications for Measurement: Improving Anonymity Online Rishab Nithyanand | Rachee Singh | Shinyoung Cho | Phillipa Gill Stony Brook University 1

  2. Anonymity on the Internet Tor Network 2

  3. Anonymity on the Internet Does not know the source Tor Network Does not know the destination 3

  4. Online Anonymity via Tor exit relay entry relay middle relay Tor Client Server 4

  5. Threat Model: Network Based Attacks Tor Network 5

  6. Threat Model: Network Based Attacks Tor Network 6

  7. Threat Model: Network Based Attacks Tor Network 7

  8. Internet routing and timing attacks AS2 AS1 AS2 AS3 AS5 AS4 AS6 AS5 Tor Network 8

  9. Internet routing and timing attacks AS2 AS1 AS2 AS3 AS5 AS4 AS6 AS5 Tor Network 9

  10. Internet routing and timing attacks AS2 AS1 AS2 AS3 AS5 AS5 AS4 AS6 AS5 AS5 Path asymmetry => Increases the attack surface [RAPTOR, USENIX 2015] TCP ACK numbers leak timing and size info on reverse path! 10

  11. Astoria [NDSS2016] + Cipollino * ● What if the Tor client could pick relays to avoid timing attacks? We show that there usually is a safe option [NDSS2016] ● Challenge: ● How can the Tor client learn network paths? ○ Astoria: Policy-based simulations on empirically derived AS graphs ● Cipollino: Based on measured paths ● 11 * https://en.wikipedia.org/wiki/Cipollino

  12. Tor client measurement-plane requirements ● Data needs to be current E.g., if a path changes to go through a new AS we need to know! ○ ● Path computations need to be local The client can’t ask a third party about paths to the destination! ○ ● Data needs to be compact + accurate ● Trade-off between: measured data as relevant as possible ( near real-time ). ○ ○ Cipollino Tor client low-latency (not on-demand measurements) 12

  13. Our solution: PathCache Basic idea: Reuse measurements already being made! ● Combine publicly accessible traceroute measurements to learn new paths ● Currently using RIPE Atlas + iPlane data ○ Augmenting with control-plane data RIPE NCC, Routeviews ○ Longer term: Efficient use of new measurements to increase coverage ○ http://pathcache.cs.stonybrook.edu ● A A A B D D B C C E E D 13 D

  14. Why is PathCache Useful? Everyone needs traceroutes! ● But why run redundant traceroutes at the expense of a constrained ● measurement budget ? Measurement hardware can be more effectively utilised. ● Standard and compact graphs as JSONs, easy to work with. ● Try it out here: http://pathcache.cs.stonybrook.edu/api/v1/174 (AS174’s dest ● based graph) http://pathcache.cs.stonybrook.edu/api/v1/path?src=2119&dst=174 ● 14

  15. Future Work? ● Need more data Huge benefit of periodic measurements from RIPE Atlas! ○ ● Path prediction as a service ● A platform to maintain AS-level paths over time (benefit of retrospective measurements) ● Provide a generic interface to upload user-run measurements. ● As a community, increase our coverage via measurement reuse 15

Recommend


More recommend