Abstract: The J?sang’s subjective trust model is studied. An evidence data preprocessing algorithm against malicious behavior is presented, experimental results indicate that trust can exponentially decrease with the increasing of the negative evidences so that malicious behavior is punished. An optimistic factor associated with entity subjectivity and system state is constructed, and transitivity operator based on the optimistic factor is presented. Theoretical analysis shows that the new transitivity operator can adapt to the system state changes. Based on evidence data normalization, new consensus operators applied to conflicting opinion are proposed. Theoretical analysis and comparison with other trust models show that these operators can improve the precision and robustness of the model, reduce the computational complexity.

Key words: trust model, trust consensus, trust transitivity, subjective logic

摘要： 研究了J?sang主观信任模型。针对模型不能惩罚恶意行为的问题，提出一种证据数据预处理算法，实验结果表明该算法能使信任随负面证据的增加呈几何级数减少，可以较好地惩罚恶意行为；针对模型传递算子不能自适应系统状态变化的问题，构建了与实体主观性和系统状态有关联的乐观因子，基于该乐观因子提出新的传递算子，理论分析表明新传递算子能够自适应系统状态变化；针对模型聚合算子聚合对立观点鲁棒性差、计算精度低的问题，基于证据数据标准化提出了新的聚合算子，理论和比较分析表明，新的聚合算子能够提高模型的精度和鲁棒性，同时降低计算的复杂性。

关键词: 信任模型, 信任聚合, 信任传递, 主观逻辑