基于贝叶斯网络和行为日志挖掘的行为信任控制

Behaviour Trust Control Based on Bayesian Networks and User Behavior Log Mining

目前评估网络用户行为的方法成本高且缺乏可操作性,为了便捷有效地对用户行为信任进行预测与评价,文中首先提出了基于贝叶斯网络的用户行为信任预测和控制算法,算法利用聚类和分布密度函数设置算法参数,建立可量化的证据与信任等级之间的对应关系.接着实现基于IIS和.Net底层架构的可配置式信任管理插件,形成用户行为日志,为预测和控制算法提供证据,免除了一般W eb日志的清洗工作.实验结果表明,文中算法可预测多属性下的行为信任等级,提高服务器的安全性和可靠性,并约束了用户的商业行为.

The existing evaluation methods of network user behaviors are of high cost and low practicability. In or- der to effectively forecast and evaluate network user behavior trust with ease, a trust forecast and control algorithm of user behaviors is proposed based on Bayesian networks, in which the clustering algorithm and the distribution density function are used to set parameters, and the corresponding relationship between quantitive evidence and trust grade is obtained. Afterwards, the configurable plug-in of trust management is implemented based on IIS and. Net framework to create user behavior logs, thus providing evidences for the forecast and control algorithm and avoiding the data cleaning of common Web logs. Experimental results indicate that the proposed algorithm is capable of predicting the trust grade in multi-trust-attribute conditions, improving the security and reliability of the server and restricting the trade behaviors of the user.