Research On Trust Mechanism Of Grid Services

WSRF-based grid platform packages various computational resources in virtual organizations of multi-agencies, provides them for the users in the form of grid services and solves the resource sharing and coordinating problems existing in the dynamic and multi-institutional virtual organizations. Grid patterns transform into an open, accessible and dynamic cooperation service mode. The grid service Quality is restricted by performance QoS and trust QoS. Performance QoS includes time-related parameters like the final period of service and accuracy-related parameters like the precision; while trust QoS is used to evaluate the reliability of the service information. However, in the current grid service, most QoS-based managements only consider the performance QoS and ignore the impact of trust QoS, or the other way around.Traditional security mechanisms (password protocols and authorization) can ensure that the grid service is immune from damage by malicious users, but they can not erase the malicious behavior caused by the lack of trust. Therefore, it is an important research subject of taking into consideration performance QoS as well as trust QoS and thus forming a grid service management strategy with trust mechanism.This paper stresses how to establish an effective trust evaluation model of grid service and fully consider the various factors influencing trust evaluation to supervise the evaluation effectively, increase the accuracy of trust evaluation and ensure performance QoS and trust QoS of grid service. The main contents are as follows:①Research on global trust model. Considering the specificity of trust in grid service and the security policy (security certificate and strategy), capability and reputation, comprehensive trust evaluation model is proposed, that is Global Trust Model based on Credential, Policy, Capability and Reputation (GTM-CPCR) to carry on a comprehensive evaluation on grid service trust. For the trust evaluation model influencing trust greatly, it also analyzes that the existing models don't take the poly-dimensional nature of trust into consideration and then proposes Reputation Computation Model base on Multi-Key Factors (RCM-MKF). In this model, many factors are taken into full consideration, such as transaction value, scoring the trust of users, time, reputation feedback score, recent trust and contribution to the community and so on.②Trust recommendation based on similarity. In the situation that there is no effective method to avoid malicious recommendation in current trust recommendation, this paper proposes Trust Recommendation Model based on Similarity. According to similarity of cosine and relative method of coefficients by Pearson, it analyzes the behavior of the two parties involved and obtains their similarity, which is then considered to be the weight of trust recommendation. On this basis, Reputation Recommendation Model Based on Similarity (RCM-MKF-Sim) is proposed for reputation feedback with direct trust. As for recommendation of indirect trust, similarity-based indirect trust computation model is proposed and the deriving method of trust path is also provided. The experiment shows that similarity-based trust recommendation model can avoid malicious recommendation behavior effectively.③Trust evaluation based on compensation. During the process of trust evaluation at present, as for the two prominent questions of heavy calculation and inaccuracy in trust evaluation, a Trust Evaluation Model based on Compensation (TEMC) is proposed. Under marketing commitment mechanism, the two parties involved in evaluation offer expected gains and compensation when failed. And the purpose of them is to maximize their own profit. According to Game Theory, during trust evaluation, the evaluated party, unclear of the evaluation reliability, can compute the reliability probability of the evaluated information according to the expected gains and compensation price proposed by the evaluating party. And then, based on the requirements for profit maximization, the recommended trade strategy which suggests two parties to take compensation as basis is derived. The evaluated party initiative of choosing the evaluating party can not only attribute to pick out the sincere evaluation effectively, but also promote to improve the accuracy of evaluation, reduce the calculation of trust degree and decrease the grid cost of conveying the evaluated information.④Qualitative research of trust model. This paper proposes qualitative-based trust model and analyzes the relationship of trust from a qualitative perspective, which can be a complement for trust relationship research or a new way of later trust research.⑤Grid service invoking based on trust. In the framework of current grid service, grid service focuses on the service finding based on functional constrains, without fully considering service filtering and choosing or providing users with enough necessary information of QoS for finishing those tasks, which can't ensure that the found services meet the users'demand on QoS. Therefore, this paper provides trust-based Service Finding (GTM-CPCR_SF) and trust-based Service Invoking (GTM-CPCR_SI) for different grid service applications. In different application environments the two algorithms can filter and choose services effectively, which can meet the demand of comprehensive QoS and performance QoS from service demanders to the maximum.