Dynamic trust model based on behavior monitoring and data mining
基于行为监控和数据挖掘的动态信任模型*

Trust relationships between entities occur in a certain context. It is difficult to use a static function to describe the complicated and variable relationship between the trust value and the multiple behavioral attributes. According to the historical behavioral data and target trust value gained by monitoring software sensors, the logistic regression analysis and pair classification method were applied to automatically perform data mining and knowledge discovery on relational schema between behavioral attributes and trust value, prior knowledge and subjective assumptions were not included, therefore, effectively solved issues on dynamic and objectivity of trust value calculation. Experimental results show that, compared with the existing model, the new model can improve the MAPE(mean absolute percentage error) and the computational efficiency and classify the trust level between entities effectively.