A Reputation System with Anti-Pollution Mechanism in P2P File Sharing Systems

摘要

File pollution has become a very serious problem in peer-to-peer file sharing systems, because of which it greatly reduces the effectiveness of systems. Users downloaded pollution files not only consumed bandwidth, but were also likely to share polluted files without checking. If these polluted files carry a virus, Trojan horse, or other malicious code, the loss of users would be disastrous. There is much research done on reputation-based anti-pollution mechanisms. Peer reputation systems and object reputation systems are two representatives reputation-based anti-pollution mechanisms currently. Peer reputation systems only care about the reputation of file providers, while object reputation systems only care about the reputation of sharing files. This paper proposes an anti-pollution mechanism combined with peer reputation and object reputation. Different from former research, we calculate the reputation of sharing files by the reputation of the voting peer. That is, object reputation is weighted by peer reputation. The reputation of the voting peer is built by its direct trust value and recommended trust value. Direct trust considers the trust of Direct Interaction Experience and the trust of Voting Similar Experience. Severe penalty strategy and voting incentive mechanism are introduced in calculating direct trust value. Therefore, a good user who uploads unpolluted files and voting on files actively can have a higher reputation, while a malicious user who uploads polluted files or voting unhonestly would have his reputation reduced sharply. Our expectation is that users for their own reputation will delect polluted files as soon as possible. These strategies give users an incentive to awareness of pollution, consequently isolate the polluters effectively. Our reputation mechanism is intended to prevent pollution spread by stimulating the awareness of users to delete files as soon as possible. At last, we simulate a P2P file sharing system to assess our reputation mechanism. The simulation results show that, compared to the object reputation system, our reputation mechanism convergence is faster, and has a better anti-pollution performance.