【作者】 赵娜；

【导师】 田俊峰；

【作者基本信息】 河北大学 ， 管理科学与工程， 2012， 硕士

【摘要】 信任主要体现在人与人交往的行为的信任度计算上，许多专家、学者越来越多的通过对电子商务商家行为的研究，信任关系的评估与异常行为的检测，来保证网络购物的安全性。由于网络商家行为的可信性具有动态性、积累性、及时性和虚假性等特点，要求我们对商家的可信评价尽可能精确真实，这需要收集相关数据，进行行为检测，并根据行为可信评价结果，对商家行为进行实时监测，保障可信评价的正确性和一致性，从而优化网络购买环境。随着电子商务在网络社会中的作用越来越凸显，人们对商家的可信性要求日益增高，商家异常监测与动态可信评价将成为商务信息安全领域研究的一个关键、难点问题。针对上述问题，本文提出了基于主观逻辑和异常行为检测的商家可信性模型（TMA-SLABD）来描述商家可信行为，商家行为由商家评价和交易信息构成，商家评价由购买者通过交易记录给商家的评分累计而成，表示商家的基本信任；交易信息则由商品信息和商家的基本属性等一系列静态信息来刻画。基于主观逻辑和异常行为检测的商家可信性模型，通过无监督异常检测方法，使用Chameleon算法对商家行为聚类分析，排除异常商家，然后根据J sang主观逻辑，改进信任计算模型解决商家动态可信问题。最后，实验结果表明本文提出的模型效率和精确性均较高。更多还原

【Abstract】 The creditability of agent is mainly embodied in terms of behaviors of interpersonalrelationships. The researching of agent behavior, the evaluation of trust relationship and themonitoring of abnormal behaviors are used more and more by many specialists and scholarsto ensure computer online shopping under the condition of security. Due to the dynamics,accumulativeness, timeliness and falsity of the credibility of agent behavior, it is required thatthe evaluation of the credibility of agent must be objective and accurate, and the trustevaluation based on agent behavior monitoring are built through collecting the relative data ofcredibility,monitoring behaviors in the running agent behavior and verifying whether theagent behavior meeting the rules of behaviors or not.As the E-commmerce plays more and more important roles in today’s network societyand the agent credibility is more and more strongly required, the dynamic trust evaluation ofagent and anomaly detection of agent behavior have become a key and difficult issue in theinformation security filed. Aiming at these problems, this paper proposes a Trust Model ofAgent Based on Subjective Logic and Anomaly Behavior Detection(TMA-SLABD). TheAgent Behavior consists of business evaluation and transaction information. Businessevaluation indicates accumulated ratings of the buyers through transaction record, which isthe common trust of agent. Transaction information is described by a series of staticinformation which consists of the commodity information and basic attributes of the agent.The Trust Model of Agent Based on Subjective Logic and Anomaly BehaviorDetection(TMA-SLABD) uses Chameleon algorithms to cluster analysis agent behaviorbased on unsupervised anomaly detection method,excluding unusual businesses. Then,theTMA-SLABD improves trusted computing model to solve the problem of agent credibilityaccording to J sang subjective logic. Finally, the accuracy and efficiency of the model isillustrated by the simulation experiment.更多还原