Understanding Pollution Dynamics in P2P File Sharing

Abstract

Pollution in P2P file sharing occurs when a large number of decoy files are injected into the P2P system. Since peers “serve” each other in the P2P file sharing system, it is obvious that pollution dynamics are closely related to user behavior. Therefore, we first conduct a human subject study to investigate user behavior. We identify the factors that are key to model user behavior, e.g., cooperativeness and awareness of pollution. Our results show that users are quite insensitive to pollution, and their behavior exhibits a bimodal distribution of the time interval between download and the quality checking of that download. We then propose a mathematical model to assess the impact of pollution on file popularity evolution. From our analysis we find that user “awareness” of pollution is a key factor in pollution dynamics. Finally, we study the impact of pollution on P2P traffic loads and show that in the worst case, pollution can quadruple the loads.