基于多属性评价的信任模型研究

发布时间：2018-05-20 14:01

本文选题：多属性评价 + 信任模型　；参考：《天津大学》2016年硕士论文

【摘要】：随着互联网技术尤其是Web 2.0的飞速发展,越来越多的web应用出现在人们的日常生活中,如P2P网络、社交网络、多Agent系统、电子商务等。随着web应用的规模不断扩大,其具有的开放性、动态性等特性使得大量的恶意用户涌现出来,给人们带来了不同程度的危害。为了尽可能地避免损失,得到令人满意的服务,研究者们提出了各种信任与信誉系统以供决策支持。本文参考国内外大量相关文献,分析了现有信任模型的优点和不足,并在此基础上,结合人类社会中信任的有关定义,提出了一种基于多属性评价的信任模型。本文的综合信任度的计算可划分为直接信任和间接信任两部分。对于直接信任,它主要来自于主体之间直接交互的经验信息,即利用用户与服务提供者交互后给予的多属性评价信息来计算。此外,由于服务提供者的行为是动态变化的,过去的行为不一能代表将来的行为,因此本文引入动态的时间遗忘因子来对历史证据进行衰减。对于间接信任,它主要来自于其他用户的推荐信息。在本文中首先获取源用户与推荐用户之间共同交互过的服务提供者的评价,接着利用欧几里得范数计算“评价对”的差异来得到推荐用户的推荐可信度,并引入一种基于证据概率分布的确定性度量方法来对推荐可信度进行动态调整,最后再利用D-S证据理论定义的证据空间到beta函数的映射来获得间接信任。在获得直接信任和间接信任后,通过引入动态权重因子,调整用户的直接信任和间接信任权重进而获得最终的综合信任度。此外,本文还考虑了恶意用户评价中的有用信息,从而为信任模型提供多样性的证据来源。本文的多属性信任模型不同于传统信任模型的单一评价,它充分利用了用户的多维度评价信息和用户偏好,从而完成对服务提供者的可信性度量。在仿真实验中,通过与经典信任模型的比较,本文的信任模型可以更好地抵御各种类型恶意节点的攻击。因此,本文的信任模型具有一定的理论意义和实际应用价值。
[Abstract]:With the rapid development of Internet technology, especially Web 2, more and more web applications appear in people's daily life, such as P2P network, social network, multi Agent system, electronic commerce and so on. With the expansion of the scale of web applications, its opening and dynamic characteristics make a large number of malicious users emerge and bring people In order to avoid losses and get satisfactory services as far as possible, researchers have proposed a variety of trust and reputation systems for decision support. This paper, referring to a large number of relevant documents at home and abroad, analyses the advantages and disadvantages of existing trust models, and on this basis, combines the relevant decisions of trust in human society. In this paper, a trust model based on multi attribute evaluation is proposed. The calculation of the comprehensive trust degree in this paper can be divided into two parts: direct trust and indirect trust. For direct trust, it mainly comes from the empirical information of direct interaction between the subjects, that is, the multi attribute evaluation information which is given by the user and the service provider is calculated. In addition, as the behavior of the service provider is dynamic, the past behavior can not represent the future behavior, so this paper introduces a dynamic time forgetting factor to decay the historical evidence. For indirect trust, it mainly comes from the recommendation information of other users. In this article, the first to obtain the source and the recommended users. The evaluation of the service providers with mutual interaction, then uses the Euclidean norm to calculate the "evaluation pairs" to get the recommended reliability of the recommended users, and introduces a deterministic measure based on the probability distribution of evidence to dynamically adjust the recommended credibility, and then use the D-S evidence theory to define the evidence space. The indirect trust is obtained by mapping the beta function. After obtaining direct trust and indirect trust, the final comprehensive trust is obtained by introducing the dynamic weight factor and adjusting the user's direct trust and indirect trust weight. In addition, this paper also considers the useful information in the malicious user evaluation, thus providing diversity for the trust model. The multi attribute trust model in this paper is different from the single evaluation of the traditional trust model. It makes full use of the user's multi-dimensional evaluation information and user preferences, thus completing the credibility measure of the service provider. In the simulation experiment, the trust model of this paper can be better offset by comparing with the classic trust model. Therefore, the trust model of this paper has certain theoretical significance and practical application value.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.08

【相似文献】