中文摘要：

针对多Agent系统中传统的信誉机制主要考虑的是事务层的信用计算，而不能很好的解决利用评价对信用计算进行扰乱和攻击的问题，设计了基于“事务信誉”、“评价信誉”的双层信誉模型和评价信誉反馈（RRF）机制．模型区分了事务信誉和评价信誉两种不同层次的信誉，利用消费方Agent的评价信誉决定对提供方Agent事务信誉的影响权重，并根据评价反馈值和评价值的距离调整以往评价者提供的评价值信誉．最后提出基于RRF机制的WS-QP算法．理论证明及实验结果表明，该模型和机制能够有效地减少恶意评价、随意评价和不公平评价等扰乱和攻击行为的影响．

英文摘要：

Mainly considering trust computing in transaction layer, traditional reputation mechanisms cannot well solve the attack and disturbance problem using rating in multi-Agent system. A double-layer reputation model based on transaction reputation and rating reputation, as well as the rating reputation feedback （RRF） mechanism, was designed. The model distinguished the two different layer reputation concepts as transaction reputation and rating reputation. The model decided the impact weight of consumer agent to transaction reputation of provider Agent using former＇s rating reputation, and adjusted past estimators＇ rating reputation based on the distance between current rating feedback value and the rating value from them. Then the weighted sum-quadratic polynomial （WS-QP） algorithm based on RRF was raised. The theor tively reduce the unfair rating. etical proof and the experimental results indicate that model and the mechanism can effecimpact of disturbance and attack problems, including malicious rating, arbitrary and unfair rating.