Mediating Tugs of War: Security-Performance Trade-offs in Privacy-Preserving Communication Systems

Abstract: In contrast to inherently compromisable client-server architectures, decentralized systems present a promising foundation for realizing anonymous and censorship-resistant communication. In addition to decentralization, relaying and delaying messages randomly is essential to protect users from surveillance and censorship. Yet, the need to obfuscate information often results in performance deficits, which prevent large-scale adoption of these systems.
In this talk, I summarize my work on performance improvements in real-world privacy-preserving communication systems and their trade-offs with anonymity and availability. In particular, I highlight my ongoing work on congestion control in Tor and my long-term involvement with the Freenet project. In the context of Freenet, I developed VOUTE, a novel design for Freenet's Friend-to-Friend overlay, which restricts application-layer connections to devices of participants sharing a real-world trust relationship. VOUTE assigns coordinates mirroring a user's position in the social graph to facilitate fast and efficient anonymous communication between any two participants. VOUTE's communication complexity scales logarithmically in the number of users, and simulations indicate that it reduces the communication overhead by more than half, at the price of revealing some information about the social graph of trust relationships.

Bio: Stefanie Roos is a post-doctoral fellow in the Cryptography, Security and Privacy group at University of Waterloo, working with Ian Goldberg. Her research focuses on privacy and security in distributed systems. In particular, she works on finding suitable trade-offs between privacy and efficiency. Currently, she aims to understand the trade-off between congestion control and traffic analysis in the Tor network with the goal of developing a secure and effective congestion control algorithm for Tor. In the past, she has worked on anonymous and attack-resistant routing schemes for P2P overlays. Some of her results have been integrated in Freenet, a P2P-based censorship-resistant publication system. Other areas of her research include Botnet detection, distributed credit networks, and next-generation Internet routing. Stefanie holds a doctorate from TU Dresden and a degree in Mathematics with Computer Science from TU Darmstadt. She was awarded the KuVS Promotionspreis 2017 honoring the best theses in networking and distributed systems in Germany for a her PhD thesis 'Analyzing and Enhancing Routing Protocols for Friend-to-Friend Overlays', supervised by Thorsten Strufe.