A multiple user sharing behaviors based approach for fake file detection in P2P environments  

作　　者：JIANG Jing, LI YongJun, FENG QinYuan, HUANG Peng & DAI YaFei Department of Computer Science and Technology, Peking University, Beijing 100871, China 

出　　处：《Science China(Information Sciences)》2010年第11期2169-2184,共16页中国科学（信息科学）（英文版）

基　　金：supported by the National Natural Science Foundation of China (Grant No. 60673183);the Doctoral Funding of MOE (Grant No. 20060001044);the National Basic Research Program of China (Grant No. 2011CB302305)

摘　　要：File pollution is a problem that is threatening security and availability in peer-to-peer environments, and could eventually be fatal to the survival of these systems. Current methods of fake file detection, which are based on trust mechanisms or file characteristics, do not consider the problems such as interference of multiple user sharing behaviors, the difficulty in obtaining large amounts of raw data, and malicious fraud immediately after version publication. This is the first time that differences in the sharing habits of users and particularly long retention time for individual files have been detected, which impact fake file detection. This paper proposes a pollution prevention model based on multiple user sharing behaviors, which reduces interference in multiple user sharing behaviors and malicious fraud immediately after version publication. It designs a low-cost implementation mechanism for use in structured peer-to-peer networks, which automatically collects a large amount of user file-sharing information and alleviates difficulties in obtaining large quantities of raw data. A parameter configuration is also given in this paper. Simulation experiments use realistic system logs and prove that this approach detects fake files accurately and quickly, thus decreasing download times of fake files, effectively guaranteeing almost 100% real file downloads and resisting file pollution.File pollution is a problem that is threatening security and availability in peer-to-peer environments, and could eventually be fatal to the survival of these systems. Current methods of fake file detection, which are based on trust mechanisms or file characteristics, do not consider the problems such as interference of multiple user sharing behaviors, the difficulty in obtaining large amounts of raw data, and malicious fraud immediately after version publication. This is the first time that differences in the sharing habits of users and particularly long retention time for individual files have been detected, which impact fake file detection. This paper proposes a pollution prevention model based on multiple user sharing behaviors, which reduces interference in multiple user sharing behaviors and malicious fraud immediately after version publication. It designs a low-cost implementation mechanism for use in structured peer-to-peer networks, which automatically collects a large amount of user file-sharing information and alleviates difficulties in obtaining large quantities of raw data. A parameter configuration is also given in this paper. Simulation experiments use realistic system logs and prove that this approach detects fake files accurately and quickly, thus decreasing download times of fake files, effectively guaranteeing almost 100% real file downloads and resisting file pollution.

关 键 词：PEER-TO-PEER file pollution user sharing behavior 

分 类 号：TP393.02[自动化与计算机技术—计算机应用技术]