WINNOWING : Protecting P2P Systems Against Pollution By Cooperative Index Filtering

Pollution (i.e., sharing corrupted files, or contaminating index information with bogus index records) is a de facto problem in many file sharing Peer-to-Peer (P2P) systems. Since pollution squanders network resources and frustrates users with unprofitable downloads (due to polluted files) and unproductive download requests (due to bogus index records), the future success of file sharing based P2P systems is questionable unless properly addressed. In this paper, we propose a novel anti-pollution scheme called winnowing. Winnowing aims to purify the index records (i.e. the information on files or the publishers) held by each index node in the system, so that download attempts based on these index records are more likely to yield satisfactory results. To attain this goal, index nodes block bogus publish messages by verifying the publisher and the contents of the publish message upon receipt of a keyword or content publish message. Second, index nodes collect feedback from the users who have downloaded files via their index records. The collected feedback is then processed and reflected in the matching index record in a novel way. Careful consideration is given to reducing the impact of false feedback, and malicious index nodes. Publish message verification has been implemented on top of the latest eMule client and extensive data has been collected from the Kad network, using this modified client. The measurement results are summarized in this paper. The the findings from the measurement study are incorporated into our analytical model, which is used to investigate the performance of user feedback mediation. The model demonstrates the effectiveness of user feedback mediation: fast convergence to near-optimal performance and insensitivity to various pollution attacks including the attacks which attempt to bypass winnowing.