摘要： 在分析信誉计算模型的基础上，提出了一种自约束的信誉更新模型，由于节点的信誉主要反映历史信誉，信誉变化主要表现在新的评价分与以前信誉的变化上，因此通过自约束因子来表达这种变化。对该模型与Bayesian模型作了仿真分析比较。结果显示：信誉评估模型体现恶意行为的惩罚思想，当一个可信节点连续出现交易失败或不可信行为时，信誉值会指数级下降，恢复其信誉则是缓慢上升。

关键词: 信任, 信誉更新, Bayesian概率模型

Abstract: Encounter or interactions between nodes unknown to each other are inevitable in Internet transaction. Reputation mechanism is to facilitate the entities trust relationship in decentralized environments. This paper provides a computational model based on directed trust and recommender trust, introduces a self-constraint reputation updating model, which focuses on the reputation rating fluctuating factor. It puts forward an idea about malicious user’s reputation punishment. By comparing with Bayesian probabilistic trust model, it shows some benefits from simulation result: in the model, reputation rating will drop off fast whenever continuous failure transactions.

Key words: trust, reputation updating, Bayesian probabilistic model

中图分类号: 

• TP309