Many anti-censorship and privacy-enhancing tools often rely on injecting sensitive data into encrypted channels that leverage popular and proprietary network services. Unfortunately, this has led to added scrutiny (or even termination) of such services by companies who disagree with the surreptitious use of their infrastructure for censorship circumvention, leaving people in information-repressive contexts with few tools to access the free and open internet.
To address this challenge, SynthMorph will develop a censorship-resistant tool that mimics user interactions with services like videoconferencing—bypassing the need for service providers’ support. The project will explore the use of machine learning techniques to synthesize traffic that inconspicuously carries covert data, so as to enable users to access the free internet while avoiding detection.
This will entail:
- training traffic-generation models based on pre-existing network traces;
- creating traffic classifiers to investigate whether the produced synthetic traces are indeed indistinguishable from real traffic; and
- developing a user-facing prototype that incorporates UX best practices.
