Dissecting Tor Bridges: A Security Evaluation of their Private and Public Infrastructures

Author(s): Srdjan Matic, Carmela Troncoso, Juan Caballero

Download: Paper (PDF)

Date: 27 Feb 2017

Document Type: Reports

Additional Documents: Slides Video

Associated Event: NDSS Symposium 2017

Abstract:

Bridges are onion routers in the Tor Network whose IP addresses are not public. So far, no global security analysis of Tor bridges has been performed. Leveraging public data sources, and two known Tor issues, we perform the first systematic study on the security of the Tor bridges infrastructure. Our study covers both the public infrastructure available to all Tor users, and the previously unreported private infrastructure, comprising private nodes for the exclusive use of those who know their existence.

Our analysis of the public infrastructure is twofold. First, we examine the security implications of the public data in the CollecTor service, identifying several pieces of data that may be detrimental for the security of bridges. Then, we measure security relevant properties of public bridges. Our results show that the 55% of public bridges that carry clients are vulnerable to aggressive blocking; that 90% of bridge clients use default bridges that are trivial to identify; that the concurrent deployment of Pluggable Transports in bridges reduces the security of the most secure transports; and that running non-Tor services in the same host as a bridge may harm its anonymity.

To study the private infrastructure, we use an approach to discover 694 private bridges on the Internet and a novel technique to track bridges across IP changes. We are first to measure the size of the private bridge population (35% discovered bridges are private) and to report existence of infrastructures that use private proxies to forward traffic to backend bridges or relays. We use a novel clustering approach to analyze the different infrastructures using proxies and bridges, examining its hosting and security properties. We provide an extensive discussion on the security implications of our findings.