Privacy and Performance in Social Overlay Networks
MetadataShow full item record
The core Internet protocols were not designed to protect the privacy of content or the identities of communicating users, thus making censorship and surveillance easy. The same problem persists at higher level protocols, popular peer-to-peer networks are trivial to monitor even for an adversary with limited resources. In fact, recent developments suggest that censorship and surveillance of Internet users is becoming more prevalent over time. This dissertation describes the design, implementation, and evaluation of two systems that can bring us towards a future where censorship and surveillance is hard. OneSwarm is a privacy-preserving data sharing network designed to give users performance com- parable to the peer-to-peer networks commonly used today but without revealing their behavior to third party surveillance. Unblock is an overlay network that leverages many of the same building blocks but is designed to help users circumvent censorship of Internet services. Common to both systems is the use of existing social trust between participants to thwart surveillance and censorship respectively. These systems are designed to run on today's Internet and require no changes to core Internet infrastructure or protocols. Measurements of the systems in the wild, and simulations of their behavior at scale, show that they protect user privacy and improve performance over existing alternatives.