Deanonymization of Tor Onion Services with Acceleration-Based Watermarking

November 25, 2022

The Tor anonymity network and its Onion Service (OS) infrastructure allow users to browse and provide services on the Internet while benefiting from sender and receiver anonymity. This enables them to circumvent censorship filters and remain safe from prosecution but also provides a means to conduct illegal activities. This has made Tor an enticing target for attackers, including Law Enforcement Agencies (LEAs) seeking to identify unlawful OSes. These LEAs may form coalitions, gaining broad access to network traffic information collected at the level of Autonomous Systems or Internet Exchange Points. This information is known to allow deanonymization attacks on Tor's infrastructure to take place. In fact, recent attacks have shown that it is possible for a global passive adversary to fully deanonymize Tor Onion Sessions with high accuracy. The proposed techniques, however, depend on the monitoring of a large amount of Tor flow samples introducing serious bottlenecks to the scalability of these attacks and to the maximum achievable recall (i.e., the number of sessions that can effectively be screened by the adversary). In this work, we propose an alternative attack, DissecTor, that utilizes a new acceleration-based watermarking scheme and machine learning techniques to deanonymize a targeted OS both effective and covertly. By identifying the target OS one significantly reduces the complexity and resources needed to fully deanonymize the whole browsing session. We performed an in depth evaluation of the DissecTor system, exploring several variants and reflecting on the best use cases for each.

Authors in the community:

• 

  A

  Afonso Manuel Vieira Machado Barros de Carvalho

  ist427000

Supervisors of this institution:

• 

  D

  Diogo Miguel Barrinha Barradas

  ist173578

• 

  N

  Nuno Miguel Carvalho dos Santos

  ist14261

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

eng - English

Instituto Superior Técnico