Improving search quality in peer-to-peer file-sharing systems

Peer-to-peer (P2P) file-sharing is one of the most significant P2P applications nowadays in terms of number of participants and volume of data transferred. The large volume and diversity of data as well as the limited search functionalities and poor file description in a P2P file-sharing system make exhaustive searches for desired content infeasible. Therefore, better ways of retrieving and ranking query results are needed.;This work introduces traditional information retrieval ranking into P2P file-sharing systems and metadata distribution techniques that are effective in finding data in query result sets, especially rare files in such a decentralized environment. Regarding the issue of poor data description, we consider the design of a fully distributed, automatic subsystem to gather other's descriptive metadata through probe queries. Experimental results indicate that the proposed techniques are effective in improving search quality with reasonable cost in both ranking and retrieving search results.;To guarantee clients find and download desired genuine files, we also work on P2P spam characterization and detection, so as to further improve the search effectiveness in P2P file-sharing systems. We propose probing technique to collect more complete feature information of query results from the network and apply feature-based ranking for automatically detecting spam in P2P query result sets. Experimental results show that the proposed techniques dramatically improve the effectiveness of P2P spam detection with reasonable cost.