点击次数：

论文类型：期刊论文

发表时间：2012-11-01

发表刊物：JOURNAL OF COMPUTER AND SYSTEM SCIENCES

收录刊物：SCIE、EI、Scopus

卷号：78

期号：6,SI

页面范围：1737-1750

ISSN号：0022-0000

关键字：Peer-to-peer file-sharing network; Reputation; Collusion; Trust community

摘要：Trust research has become a key issue in the last few years as a novel and valid solution to ensure the security and application in peer-to-peer (P2P) file-sharing networks. The accurate measure of trust and reputation is a hard problem, most of the existing trust mechanisms adopt the historical behavior feedback to compute trust and reputation. Thus exploring the appropriate transaction behavior becomes a fundamental challenge. In P2P system, each peer plays two roles: server and client with responsibility for providing resource service and trust recommending respectively. Considering the resource service behavior and trust recommending behavior of each peer, in this paper, we propose a new trust model adopting the technology to calculate eigenvectors of trust rating and recommending matrices. In our model, we define recommended reputation value to evaluate the resource service behavior, and recommending reputation value to evaluate the trust recommendation behavior. Our algorithm would make these two reputation values established an interrelated relation of reinforcing mutually. The normal peers provide authentic file uploading services, as well as give correct trust recommendation, so they can form a trusted and cooperative transaction community via the mutual reinforcement of recommended and recommending reputation values. In this way, the transaction behaviors of those malicious peers are isolated and confined effectively. Extensive experimental results also confirm the efficiency of our trust model against the threats of exaggeration, collusion, disguise, sybil and single-behavior. (C) 2011 Elsevier Inc. All rights reserved.