TESLA-Based Defense against Pollution Attacks in P2P Systems with Network Coding

Pollution attacks are well-known to have detrimental effect on intra-session network coding in general, and in peer-to-peer (P2P) systems with network coding in particular. Previously proposed defense mechanisms against pollution attacks in intra- session network coding face various challenges that make them ill-suited for P2P systems: they are either computationally expensive, not collusion resistant, or work only on fixed, known topologies. In this work, we propose a novel, complete defense system for network coding-based P2P systems that can (i) quickly detect corrupted blocks, (ii) precisely identify the attackers, thereby eliminating them from the network, (iii) resist arbitrary collusion, and (iv) work with unknown, dynamic topologies, as it is the case in P2P systems. Our scheme uses and builds on two key ingredients: homomorphic message authentication codes and time asymmetry (as in TESLA) to provide source authentication for the detection scheme and non-repudiation for the identification scheme. Our mechanisms introduce significantly less communication and computation overhead than other comparable state-of-the-art schemes for P2P systems. Using implementation in both C/C++ and Java, on both a PC and an Android device, we show that the computational delay per block at each peer is as low as 600 microseconds and the bandwidth overhead is as low as 1.3%.