【作者】 徐军；

【导师】 许峰；

【作者基本信息】 南京航空航天大学 ， 软件工程， 2016， 硕士

【摘要】 随着“互联网+”时代的到来,电子商务已经成为新型经济驱动发展的关键部分。电子商务所发生的安全危机将直接造成用户的经济损失,并且泄露用户的个人隐私的信息,直接危害到电子商务的各个参与方。因此,为保证电子商务在不同交易环境下的交易可靠、安全及确保服务质量,如何准确定量表示和预测信任关系以增进交易双方的信任问题都是亟待解决的。面对越来越严峻的电子商务安全形势,本文对电子商务中存在的安全问题进行了分析,总结了以信任管理为中心的电子商务安全体系框架,阐述了上下文信任管理模型中的关键技术。由于电子商务环境下的信任机制的评估问题相较于传统的P2P环境更为复杂,为提高信任评估的准确性和效率,我们以电子商务环境中的上下文因素和信息节点的选取为重要研究对象,并在此基础上构建信任模型和评估方法。本文的主要研究贡献如下:首先,针对电子商务中的信任评估问题,设计了一种基于上下文多因素的模糊综合评价模型,能够动态调节信任计算的权重。该模型充分考虑了影响电子商务信任的上下文因素,以用户上下文、交易上下文、环境上下文为主因素进行建模,通过对评估用户前一阶段历史信任评价值的准确性做出判断,进一步优化权重的动态调节。最后,通过仿真实验进行了对比分析,结果表明本文提出的CFTrust信任模型表现出了良好的计算准确度,能够有效抑制恶意节点的攻击。其次,针对信任模型中的恶意攻击问题,通过考虑信任自身的时间衰减及在电子商务中交易金额对信任的影响,提出了基于时间衰减和交易金额的多维信任评估模型。通过使用动态的时间滑动信任预测算法和交易金额产生信任的动态波动算法,能有效抵抗恶意攻击行为。通过对不同恶意攻击策略的仿真实验,结果显示本文所优化的CFTrust信任模型具有较低的均方根误差和良好的抗攻击能力。最后,针对信任计算过程中推荐节点选取效率较低的问题,提出了一种基于概率分布映射的推荐节点优化选取算法,与局部广播的推荐节点选择方法在推荐信息获取的效率和准确性方面做了仿真对比分析。实验证明,基于概率分布的推荐节点选择算法能够更快速准确的选取推荐节点,准确计算目标节点的信任值,为网络高效运行提供保障。更多还原

【Abstract】 With the advent of the "Internet plus" era, e-commerce has become a key part of the development of new economic driven. The security crisis of electronic commerce will directly result in the economic loss of users, and the disclosure of the user’s personal privacy information will directly harm to the electronic commerce of all participating parties. Because of the openness and anonymity of the Internet, trading environment and entity are dynamic and unpredictable. Transactions are often carried out between completely strange entities, which greatly increase the risk of security and uncertainties faced by both parties. Therefore, in order to ensure the reliability, security and quality of service of the electronic commerce in different trading environment, how to accurately and quantitatively express and predict the trust relationship for the sake of enhancing the trust of both sides of the transaction is to be resolved.Facing more and more severe e-commerce security situation, this thesis analyses the existing security problem in the electronic commerce, summing up the framework of electronic commerce security system based on trust management, and explaining the key technologies of trust management model in the context. Due to the evaluation of trust mechanism is more complex under the electronic commerce environment compared with the traditional P2 P environment, for improving the accuracy and efficiency of trust evaluation, we build a trust model and assessment methods on the basis of taking the context of the e-commerce environment and the selection of information peers as important research objects.The research work of this thesis is mainly reflected in the following aspects:Firstly, a fuzzy comprehensive evaluation model is designed based on the context of multi factor, which can dynamically adjust the weights of trust calculation. Fully consider the context factors that influence the trust of e-commerce, the user context, transaction context, the environment of the main factors of the model, through the evaluation of the user’s first phase of the accuracy of the value of the trust evaluation, and further optimize the weight of the dynamic adjustment. A comparative analysis is carried out through the simulation experiments, the results show that the CFTrust trust model proposed in this thesis is a good accuracy, and can effectively inhibit the attack of malicious peers.Secondly, for the problems of malicious attack in trust model, an anti-attack trust model based on time attenuation and transaction amount is proposed by taking into account of the time decay of trust itself and the influence of the transaction amount in the electronic commerce. By using dynamic time slip trust prediction algorithm and the dynamic volatility algorithm of transaction amount, it can effectively resist the malicious attacks. Through the simulation experiment of different malicious attacks, the results show that the optimized CFTrust trust model has a low root mean square error and good anti-attack capability.In the end, the thesis proposes one kind of the recommended peers selected optimization algorithms based on the probability distribution of mapping, which is aiming at the low efficiency problem of recommended peer selection in the process of trust computation, whose efficiency and accuracy of the proposed method are compared with the local broadcast. The experimental results show that the recommendation algorithm based on the probability distribution can be used to select the recommendation peers more quickly and accurately, and accurately calculate the trust value of the target peer, and provide guarantee for the efficient operation of the network.更多还原