Mid-defense: Mitigating protocol-level attacks in TOR using indistinguishability obfuscation

摘要

TOR is one of the most famous anonymity networks. TOR works correctly when at least one honest relay exists along an established circuit. However, many attacks reveal anonymity of TOR communications by watermarking traffics, changing timings, counting cells per packets, and recently, introducing errors in cells. This paper focuses on protocol-level attacks which can de-anonymize users by changing a cell, and provides a solution for this attack by changing the way integrities of cells are checked. The proposed design allows all relays, not just the exit node, to check the integrity of cells, independently. In addition, this paper proposes a novel method, based on the concept of indistinguishability obfuscation, for dummy traffic generation within the middle relay. Finally, a method is introduced to distinguish between real and dummy flows, solely by an onion proxy (on a client's machine).