Research On Multi-dimension Trust Vector And Double-layer Trust Model Based Trust Architecture In Multi-agent Distributed Cooperation Environment

With the network nodes developing in intelligent and service direction in distributed computing environment, Intelligent Agent (Intelligent Agent, IA) and Multi-agent System (Multi-Agent System, MAS) become important technologies to support new distributed cooperation computing with open, reconfigurable and scalable property.In the multi-agent system, the trust and reputation mechanism will help intelligent agent find trustworthy and reliable service collaborators, ensure distributed cooperation progressing nomaly and efficiently, and play a positive role to agent managing the distributed system of increasingly complex and expansion.In recent years, the researchs on the trust and reputation mechanism have developed many trust and reputation models based on different principles and contexts. However, the current trust and reputation models in MAS distributed cooperation system can not solve problems about unfair evaluation, low-motive evaluation, collusion and personal requirement of trust, and thus affected the practical application of trust and reputation mechanism. Therefore, this thesis researches Vector-based Personal Decision Double-Layer Trust Architecture, VBPDDLTA, in MAS, established and improved the basic priciples and technology framework of the system.Aiming at the problem of traditional single-layer trust models cannot solve the disturbance and attackment to trust computing from incorrect rating information well, such as unfair rating, random rating and collusion, this thesis presents Double-Layer Trust Model, DLTM. The Model distinguishes two different property trust concepts, Transaction Trust, TT and Rating Trust, RT, presents double-layer reliability framework based on transaction-layer and rating-layer, and dynamicly maintains the trustworth of tranction and rating service relying on the relation of inter-dependence and effection between two layers. The experiment proved this model not only could solve traditional trust problems, but also could deal with the attacking actions to rating system.Aiming at the problem of the traditional computing style based on single trust value can not express multi-dimensional meaning of trust in many applications, can not richly and accurately show the rators' reliability status, and not support trust decision related to application domain and personal trend, this thesis presents trust expression based on multi-dimensional trust vector. The method of replacing traditional single-value trust with multi-dimensional vector trust can make each dimension of trust vector accurately, so as to support agent expressing personal requirement and decision judgement about service. The results proved that the introduction of multi-dimensional trust vector improved agent's correctness of making personal decision. Aiming at the problem of traditional trust decision mechanism based on single value trust normally using simple threshold-based judgement couldn't meet complex requirement in application, this thesis presents trust vector based personal decision mechanism. With the supporting of multi-dimension trust model, this mechanism attains detailed classification of reliability vector, combines with the factors of application-related and personal trend, and computes reliability value with richly personal information, so as to get more effective trust decision. The result proves that multi-dimensional trust vector based personal decision mechanism improves the correctness of agent selecting appropriate service provider, and solves the problem of personal requirement.Finally, this thesis presents Vector-based Personal Decision Double-Layer Trust Architecture, VBPD_DLTA. This model combines Vector-based trust model with personal decision layer, introduces system's architecture, internal components, and mutual interaction. Experiments have proved that the system is feasible.Comparing with current trust systems, Vector-based Personal Decision Double-Layer Trust Architecture, VBPD_DLTA, can effectively solve the problems of attacking and disturbing trust computing, such as unfair rating, random rating and collusion, meet the needs to application-related and personal trend based trust decision, and enhance trust system's robustness and practicality.