【作者】 许光全；

【导师】 冯志勇；

【作者基本信息】 天津大学 ， 计算机应用技术， 2007， 博士

【摘要】 目前,计算机和Internet网络已经渗透到社会生活的各个方面,人们越来越多地依赖于计算机,依赖于网络。同时,随着计算机网络应用的深入,安全问题越来越突出。最为明显的是出现了虚拟社会中由于信任危机带来的交易安全隐患。如何缓解这种信任危机,确保基于信任的交易安全,已经成为当前必须解决的难题。本文从社会认知的角度出发,研究信任构建和社会关系网络分析,建立虚拟社会中信任评价和管理的可计算模型,所取得的主要研究成果如下:(1)详细分析了信任的认知结构和社会关系网络,给出了若干有关信任研究的前提假设,提出基于“三方互惠决定论”的信任社会关系网络的形式化描述。同时探讨了非认知性信任和情感性信任的基本结构,提出了虚拟社会中信任管理系统的概念模型。该模型将信任凭证和信任策略有效地结合起来,基于模块化的思想,给出整个系统信任评价策略,用户可以根据实际需要灵活地选用不同的模块组成不同的信任管理系统。(2)针对虚拟临时系统环境下快速信任的特殊性,分析和研究了快速信任的产生机制,建立起快速信任推理的概念模型。在此基础上,研究和扩展了Dempster -Shafer理论在信任建模中的应用,将快速信任量化并建立可计算模型,并通过一个实例详细阐述了该模型的应用。理论计算和仿真实验表明,该快速信任的证据理论模型是有效的。(3)分析研究了科尔曼信任模型的改进方法学,利用非齐次马尔可夫系统(FNHMS)模型来模拟多agent系统,提出了多agent系统中信誉的管理方法。该方法通过设置母体参数和基本参数的方法来得到保持转移概率的随机性,进而得到了整个多agent系统的基于信誉的分布情况。最后在实验过程中,通过改变参数的方法来模拟不同信任评价者的信任偏好,得到了不同的信任偏好对整个系统的信誉分布的影响。综上所述,通过对认知性信任、非认知性信任以及情感性信任的结构成分分析,建立了一个模块化的虚拟社会中信任管理的概念模型。然后研究了虚拟临时系统中快速信任的产生机制,并建立了基于证据理论的快速信任推理模型。最后,采用模糊非齐次马尔可夫系统来模拟多agent系统的信誉分布情况,并通过改变参数的方法研究了不同信任偏好对整个系统信誉分布情况的影响。该研究有助于人类更好的认识和掌握虚拟社会中的信任机制,构建一个高可信的网络交易环境,确保虚拟社会的网络安全。更多还原

【Abstract】 Nowadays computer and Internet have been popular in every walk of life, people are becoming more and more dependent on computer and network. In the meanwhile, with the application of computer network, security problem becomes more serious. One of the obvious challenges is trust crisis and its influence on interaction security. How to alleviate the trust crisis and guarantee the security during interaction in virtual society has become the urgent problem. This paper starts from the social cognition discipline to research trust ingredients and social network analysis, and then build up the computational model on rating and management for trust and reputation. The main work includes the following:(1) Cognitive ingredients and social network for trust are analyzed, then several assumptions about trust research are given, the formal description for social network based on so-called“the mutual decision of the tripartite”is realized. Furthermore, ingredients for non-cognitive trust and affective trust are researched, based on which the conceptual model on trust management is advanced. This model combines trust certifications and trust strategies, and is mainly based on the modular thinking, so that users can freely combine different modulars into a new trust management system, which has the characterization of flexibility.(2) As for the unique characteristics of trust in virtual temporary system, the origins for the swift trust are researched and then a conceptual model is built up, based on which the Dempster-Shafer theory of belief functions is extended to be used in trust modeling. Then a computational model based on D-S theory is built up, and an example is run to expatiate on the application. Lastly the following theoretic computing and experiments in NetLogo verify its correctness and fitness.(3)After the analysis of improvement methodology of Coleman model of trust, FNHMS is used to manage reputation in multi-agent systems. In this method, to maintain the stochastic property of the transition probability matrix, population parameters and basic parameters are introduced, and then the distribution of agents is gotten. Lastly by altering the value of index, the influence on reputation is researched when trust preferers are changed.In a word, firstly the ingredients of cognitive trust, non-cognitive trust and affective trust are investigated, based on which a conceptual model about trust management is built up in virtual society. Then the swift trust mechanism is researched in virtual temporary system, next a computational model of swift trust is built up based on D-S theory. FNHMS is exercised to simulate the multi-agent systems, in this way the management of agents about reputation is realized. Finally, by altering the parameter index, the change of reputation for different trust preferers is simulated. Such research in this paper is helpful for human to master the trust mechanism, construct a trustable interaction environment, and guarantee the network security in a virtual social.更多还原