Mistrustful P2P: Privacy-preserving File Sharing Over Untrustworthy Peer-to-Peer Networks

Abstract

Peer-to-Peer networks are extensively used for largescale file sharing. As more information flows through these networks, people are becoming increasingly concerned about their privacy. Traditional P2P file sharing systems provide performance and scalability at the cost of requiring peers to publicly advertise what they download. Several P2P privacyenhancing systems have been proposed but they still require peers to advertise, either fully or partially, what they download. Lacking alternatives, users have adopted anonymity systems for P2P file sharing, misunderstanding the privacy guarantees provided by such systems, in particular when relaying traffic of insecure applications such as BitTorrent. Our goal is to prevent any malicious peer(s) from ascertaining users' content interests so that plausible deniability always applies. We propose a novel P2P file sharing model, Mistrustful P2P, that (1) supports file sharing over open and untrustworthy P2P networks, (2) requires no trust between users by avoiding the advertisement of what peers download or miss, and (3) still ensures deterministic protection of user's interests against attacks of size up to a configured privacy protection level. We hope that our model can pave the ground for a new generation of privacyenhancing systems that take advantage of the new possibilities it introduces. We validate Mistrustful P2P through simulation, and demonstrate its feasibility.