TorKameleon: Improving Tor's Censorship Resistance With K-anonimization and Media-based Covert Channels

The use of anonymity networks such as Tor and similar tools can greatly enhance the privacy and anonymity of online communications. Tor, in particular, is currently the most widely used system for ensuring anonymity on the Internet. However, recent research has shown that Tor is vulnerable to correlation attacks carried out by state-level adversaries or colluding Internet censors. Therefore, new and more effective solutions emerged to protect online anonymity. Promising results have been achieved by implementing covert channels based on media traffic in modern anonymization systems, which have proven to be a reliable and practical approach to defend against powerful traffic correlation attacks. In this paper, we present TorKameleon, a censorship evasion solution that better protects Tor users from powerful traffic correlation attacks carried out by state-level adversaries. TorKameleon can be used either as a fully integrated Tor pluggable transport or as a standalone anonymization system that uses K-anonymization and encapsulation of user traffic in covert media channels. Our main goal is to protect users from machine and deep learning correlation attacks on anonymization networks like Tor. We have developed the TorKameleon prototype and performed extensive validations to verify the accuracy and experimental performance of the proposed solution in the Tor environment, including state-of-the-art active correlation attacks. As far as we know, we are the first to develop and study a system that uses both anonymization mechanisms described above against active correlation attacks.