Research On Anti-Content Pollution Mechanism In P2P File Sharing System

With the rapid growth in the number of users and shared files, P2P file sharing system becomes one of the most important applications in today's Internet. P2P file sharing system possesses a lot of good characteristics, such as good load balance and scalability. But as consequence of their distribution, openness and autonomy, P2P file sharing systems have been targets of several malicious user behavioral patterns, such as file pollution and index pollution.In the file pollution attack, attackers in P2P file sharing system corrupt the targeted content, rendering it unusable, and then make the polluted content available for sharing in large volumes. In this manner, polluted files spread through the file sharing systems. File pollution not only reduces the usability of the P2P file sharing systems but also provides an ideal platform for the propagation of virus and worms. In the index pollution attack, attackers insert a lot of bogus index records and disturb the downloading behavior of users. If there are many bogus records pointing to the same node, the node will suffer severe Denial-of-Service (DoS) attack. In this dissertation, both file pollution and index pollution are regarded as content pollution. And we will focus on modeling pollution in P2P file sharing systems and proposing effective anti-pollution mechanism.The existing research on pollution propagation model does not consider the index pollution. But in real P2P file sharing system file pollution and index pollution actually influence each other and impact the file propogation finally. So this dissertation proposes a unified model of file pollution and index pollution. The unified model classifies nodes'state into five categories, and analyzes the co-effect of the file pollution and index pollution. We also quantitatively evaluate the effect of pollution with two metrics which are the abort ration of peers who quit with no unpolluted files downloaded and the average download time of an unpolluted file.As the prevalence of content pollution in P2P file sharing system, there is a demand for cost-effective solutions to isolate polluters and minimize pollution dissemination. This dissertation proposes R&Ptrust, an anti-pollution mechanism based on resource and peer reputation, where reputations of peers and resources are computed globally. Moreover, R&Ptrust imposes severe and quick punishment to active polluters, while also includes incentive mechanism for passive polluters to delete the polluted resources in their sharing folders. We evaluated R&Ptrust by comparing its efficiency in reducing pollution dissemination against the previously proposed Peertrust and Filetrust. Our results show that R&Ptrust has a much faster convergence to a competitive efficiency and keeps good anti-pollution performance even under collusion and slander attack.