TORKAMELEON. IMPROVING TOR’S CENSORSHIP RESISTANCE WITH K-ANONYMIZATION MEDIA MORPHING COVERT INPUT CHANNELS

• Autor(a) principal: Vilalonga, João Afonso Teixeira Mendes
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/151105
• topic: Privacy-Preserving Communication; Anonymous communication; Tor Network; Tor Deanonymization Attacks; Traffic Correlation; Global Censorship Adversaries; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

Anonymity networks such as Tor and other related tools are powerful means of increas- ing the anonymity and privacy of Internet users’ communications. Tor is currently the most widely used solution by whistleblowers to disclose confidential information and denounce censorship measures, including violations of civil rights, freedom of expres- sion, or guarantees of free access to information. However, recent research studies have shown that Tor is vulnerable to so-called powerful correlation attacks carried out by global adversaries or collaborative Internet censorship parties. In the Tor ”arms race” scenario, we can see that as new censorship, surveillance, and deep correlation tools have been researched, new, improved solutions for preserving anonymity have also emerged. In recent research proposals, unobservable encapsulation of IP packets in covert media channels is one of the most promising defenses against such threat models. They leverage WebRTC-based covert channels as a robust and practical approach against powerful traf- fic correlation analysis. At the same time, these solutions are difficult to combat through the traffic-blocking measures commonly used by censorship authorities. In this dissertation, we propose TorKameleon, a censorship evasion solution de- signed to protect Tor users with increased censorship resistance against powerful traffic correlation attacks executed by global adversaries. The system is based on flexible K- anonymization input circuits that can support TLS tunneling and WebRTC-based covert channels before forwarding users’ original input traffic to the Tor network. Our goal is to protect users from machine and deep learning correlation attacks between incom- ing user traffic and observed traffic at different Tor network relays, such as middle and egress relays. TorKameleon is the first system to implement a Tor pluggable transport based on parameterizable TLS tunneling and WebRTC-based covert channels. We have implemented the TorKameleon prototype and performed extensive validations to ob- serve the correctness and experimental performance of the proposed solution in the Tor environment. With these evaluations, we analyze the necessary tradeoffs between the performance of the standard Tor network and the achieved effectiveness and performance of TorKameleon, capable of preserving the required unobservability properties.