基于情感强度感知的动态信任评估机制研究

发布时间：2018-08-26 14:57

【摘要】：近年来,随着互联网的兴起,开放计算系统(Agent系统、P2P网络、Ad Hoc网络等)和电子商务(eBay、Amazon等)等应用得到了快速发展。然而,其自身的开放性、匿名性、欺骗性等动态特性使得各种网络安全问题日益突出,安全性问题已成为制约其发展的关键性因素。而信任评估机制为开放式网络环境下的安全提供了一种保障方案,因此如何构建合理有效的信任评估机制是一项重要任务。对于信任评估结果的准确性来说,现有的信任评估机制还存在一些可以改进的地方,如普遍未充分考虑信任的不确定性问题,忽略了对信任的否定及犹豫信息的描述,未考虑到文字评论和等级评价的不一致问题,不能满足用户对评估准确性的要求等。因此,本文提出了一种基于情感强度感知的动态信任评估机制。该模型将信任管理与情感分析技术相结合,对信任的表达方式用情感强度向量重新给出了定义,根据信任属性数据对实体间的动态信任关系进行评估,从而得到对实体合理的评估结果。首先,收集用户终端的信任评论信息,评论信息是用户对信任多维属性评价的组合。利用对象抽取技术进行信任属性和对应情感词的抽取,并根据中英文情感词典计算情感词在各极性(正面、负面和中性)的情感强度值,实现了信任属性评价与情感强度向量的对应关系。从而更加合理和完备地对信任进行了形式化描述,考虑了信任的不确定性,消除了用户真实评价与等级评价不一致的问题。其次,在量化后的信任评价数据基础上,以信任属性数据为计算单位设计信任计算算法,算法中充分结合了用户间的历史交互数据和其他用户给出的推荐信任数据,计算出服务实体的信任值。考虑到信任随时间衰减的动态特性,通过指数分布法为不同时期的用户历史交易分配时间权重,从而使得对直接信任的评估更符合服务实体的近期表现,实现了对行为不稳定的策略型恶意攻击的有效遏制。在反馈信任度计算中,给出了反馈可信度的计算方法,实现了对虚假、欺骗的恶意反馈评价的有效鉴别。然后,联合得分函数和精确函数比较用户总体信任度,给出了基于情感强度向量的服务实体选择策略。最后,本文基于以上理论,设计并实现了基于情感强度感知的信任评估机制,并通过仿真实验对信任评估机制的性能进行了研究。实验结果表明所提模型能够有效描述实体间的信任关系,显著识别并消除多种恶意攻击行为,提高了信任评估的准确性。
[Abstract]:In recent years, with the rise of the Internet, the applications of open computing system (Agent system, P2P network, Ad Hoc network, etc.) and electronic commerce (eBay,Amazon etc.) have been developed rapidly. However, its dynamic characteristics, such as openness, anonymity and deceptiveness, make all kinds of network security problems stand out day by day, and security problems have become the key factors restricting its development. The trust evaluation mechanism provides a guarantee scheme for the security in the open network environment, so how to construct a reasonable and effective trust assessment mechanism is an important task. For the accuracy of the results of trust assessment, there are still some improvements in the existing trust assessment mechanism. For example, the uncertainty of trust is not fully considered, and the description of negative and hesitant information about trust is ignored. It does not take into account the inconsistency between text comments and rating evaluation, and can not meet the requirements of users for the accuracy of evaluation and so on. Therefore, this paper proposes a dynamic trust evaluation mechanism based on emotion intensity perception. This model combines trust management with emotion analysis technology, and redefines the expression of trust with emotion intensity vector, and evaluates the dynamic trust relationship between entities according to the trust attribute data. Thus, the reasonable evaluation results of the entity are obtained. Firstly, the trust comment information of user terminal is collected, which is a combination of users' evaluation of multi-dimensional attributes of trust. The object extraction technique is used to extract the trust attribute and the corresponding affective words, and the emotional intensity of the affective words in all polarity (positive, negative and neutral) is calculated according to the Chinese and English emotional Dictionary. The relationship between the evaluation of trust attribute and the emotion intensity vector is realized. Therefore, the formal description of trust is more reasonable and complete, the uncertainty of trust is considered, and the problem of inconsistency between user's real evaluation and grade evaluation is eliminated. Secondly, on the basis of quantized trust evaluation data, trust computing algorithm is designed with trust attribute data as the unit of calculation. The algorithm fully combines the historical interactive data between users and the recommended trust data given by other users. Calculates the trust value of the service entity. Taking into account the dynamic characteristics of trust decay over time, the exponential distribution method is used to assign time weights to users' historical transactions over different periods, thus making the assessment of direct trust more consistent with the recent performance of service entities. An effective containment of strategic malicious attacks with unstable behavior is achieved. In the calculation of feedback trust, the calculation method of feedback credibility is given, and the effective identification of false and deceptive malicious feedback evaluation is realized. Then, the total trust degree of the user is compared with the precision function, and the service entity selection strategy based on the emotion intensity vector is given. Finally, based on the above theory, a trust evaluation mechanism based on emotion intensity perception is designed and implemented, and the performance of trust evaluation mechanism is studied through simulation experiments. The experimental results show that the proposed model can effectively describe the trust relationship between entities, identify and eliminate many malicious attacks, and improve the accuracy of trust evaluation.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.08

【相似文献】