一种正态分布下的动态推荐信任模型  

Normal Distribution Based Dynamical Recommendation Trust Model

作　　者：邵堃[1] 罗飞[1,2] 梅袅雄[1] 刘宗田[3] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009 [2]中国邮政储蓄银行安徽省分行科技发展部,安徽合肥230031 [3]上海大学计算机工程与科学学院,上海200072

出　　处：《软件学报》2012年第12期3130-3148,共19页Journal of Software

基　　金：国家自然科学基金(60975033,60575035,60275022,69985004);安徽高校省级自然科学研究重点项目(KJ2010A272)

摘　　要：建立基于连续过程的推荐信任模型,描述间接信任这种最复杂的信任关系,在保障开放环境的安全和开放系统的可靠运行方面有着重要意义.通过量化间接信任影响因素,运用分级剪枝方法过滤推荐信息,将结果作为正态过程的采样样本,计算获取后验分布期望的Bayesian估计值.在此基础上,详细阐述信任动态演化的过程,深入探讨信任度和可信度之间的关系,给出了命题及其数学证明.实验数据表明,模型提高了抵御恶意攻击的能力,得出了更加有效和精确的结果,与相关命题的数学推导相一致.On the basis of a continuous process, the recommendation trust model is built for depicting indirect trust, the most complicated trust relationship. The model is also significant for the security and reliability of an open environment and open systems. After the factors influencing indirect trust are quantified, the filtrated recommendations are regarded as samples from the normal process, and the Bayesian estimation value of posteriori distribution expectation is obtained from calculation. Next, after an elaborate discussion of the trust evolution and the relationship between trust value and trustworthiness along with some propositions and proofs are proposed. Experimental data show that with the capacity of resisting malicious attacks improved, the model gives a more effective and precise result, which is also consistent with mathematical deduction.

关 键 词：推荐信任模型 Bayesian估计 正态分布 可信度 收敛性 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]