Peer-to-Peer Anonymous Networking Using Coding

In this paper, we consider design and analysis of coding-based anonymous routing systems in peer-to-peer (P2P) overlay networks. An unknown subset of participating nodes is adversarial, and can collude to try to identify the communicating nodes through passive observations. The first part of this work considers subgraph setup in the absence of a reliable public key infrastructure (PKI). As in the "slicing the onion" scheme proposed by Katti et al., a sender constructs a layered subgraph over which coding is performed, but we propose a new coding scheme with a formal information theoretic security characterization. We consider optimization of protocol parameters to maximize the adversary's uncertainty, as measured by the entropy of the source and sink identities, and show that a randomized strategy can improve anonymity and resource usage efficiency. The second part of this work focuses on the data transmission phase, assuming availability of a subgraph setup scheme (either PKI-based or coding-based) and end-to-end encryption. We use network coding at intermediate nodes to improve networking performance and reduce complexity by replacing expensive cryptographic operations at each hop with simpler linear algebra operations.