Anonymous Networking against Active Adversaries
- Tracey Ho, Assistant Professor of Electrical Engineering and Computer Science, California Institute of Technology
- Michelle Effros, Professor of Electrical Engineering, California Institute of Technology
We consider anonymous communication in a peer-to-peer (P2P) system in the presence of active adversarial nodes whose identities are a priori unknown. Existing work has proposed network coding strategies for anonymous communication over rectangular overlay subgraphs where the adversaries seek to identify source and sink nodes. In this work we further consider active attacks by adversarial nodes seeking to disrupt communication through the network. We propose a communication protocol in which subgraphs are periodically changed based on reverse feedback paths from the sink as to whether valid or corrupted information is received. The source nodes analyze the network and deduce adversarial likelihoods for nodes in the network based on previous communication sessions. From this analysis, effectiveness of attacker nodes is reduced through fewer appearances in future subgraph choices. We determine a practical approach for network selection based on performance analysis and anonymity considerations, with performance measured by throughput of information over a given timeframe.