Research On Trust Mechanism In Virtual Communities

With the development of information technology and Internet applications, the research of network security is facing new challenges. Open network environment means not only the existence of large amount of dynamic users, but also high possibility of interactions between unknown entities. Because the entities maybe not belong to the same administration domain, some traditional security technologies such as identity authentication, authorization or cryptography are not applicable. Trust, as a promising method of security technologies for open network, attracts more and more attention these years.Virtual Community is a significant application in the Internet. With the exponential growth of the virtual community, more and more studies about virtual community have been conducted in sociology and economics. Researches show that an effective trust mechanism is crucial to fertilize the development of the virtual community.As a kind of socialized network, the trust mechanism of virtual community is different from other applications. In computer science field, there are not enough studies on virtual community and relevant trust mechanism. The research of this dissertation focuses on the trust model design, relevant algorithm and dynamic policy management in virtual community.The main research contents of this dissertation are as follows:1) Analyzing the subjectivity, uncertainty, measurability, context-dependent, dynamic, asynchronies of trust, and pointing out that the focus of network trust is on the provision trust. After giving an overview of existing and proposed systems, some disadvantages are concluded.2) Analyzing some dominant issues in the trust model design of virtual community, and giving some design principals. A novel trust model of virtual community-CommuTrust is proposed. CommuTrust differentiates reputation from trust, and evaluate provision trust and recommendation trust respective. Three basic attributes: reputation, recommendation trust and active strength are used to indicate the status of the community members. CommuTrust, which has little cost of storage and traffic, is suitable to virtual community with high dynamic and heterogeneity. 3) Proposing a new bi-rating based rating verification scheme, which are helpful to alleviate the affection from malicious members and unfair ratings. The rating verification Scheme gives good incentive to rating submission, fair rating and identity consistency. The experiments show that this scheme can improve the performance of the trust model, and works well even in severe environment.4) Proposing a control theory based updating algorithm for recommendation trust. This algorithm is simple but effective. The experiments show that it has good performance for recommendation trust assessment, and can identify malicious members quickly.5) Proposing new trust and reputation aggregation algorithms. The trust integration algorithm presents the subjective, time-dependent natural of trust, while the reputation aggregation algorithm presents its history-dependent stability and fragility. Both algorithms are helpful to the development of virtual community.6) Analyzing the features of trust in three different phases of trust formation and giving some suggestions for the dynamic trust policy management in virtual community.