Cecylia
Bocovich,
PhD
candidate
David
R.
Cheriton
School
of
Computer
Science
The increasing centralization of Internet infrastructure and web services, along with advancements in the application of machine learning techniques to analyze and classify network traffic, have enabled the growth and proliferation of Internet censorship. While the Internet filtering infrastructure of censoring authorities improves, cracks and weaknesses in the censorship systems deployed by the state allow Internet users to appropriate existing network protocols in order to circumvent censorship attempts. The relationship between censors and censorship resistors is often likened to a cat-and-mouse game in which resistors struggle to find new gaps in nation-state firewalls through which they can access content freely, while censors are devoted to discovering and closing these gaps as quickly as possible.
The life cycle of censorship resistance tools typically begins with their creation, but often ends very quickly as the tools are discovered and blocked by censors whose ability to identify anomalous network traffic continues to grow. In this thesis, we provide several recipes to create censorship resistance systems that provably disguise user traffic, despite a censor's complete knowledge of how the system works. We describe how to properly appropriate protocols, maximize censorship-resistant bandwidth, and deploy censorship resistance systems that can stand the test of time.