基于D-S证据理论的可信评估机制研究

发布时间：2018-03-21 18:29

本文选题：开放网络　切入点：信任评估　出处：《南京邮电大学》2014年硕士论文　论文类型：学位论文

【摘要】：近年来，随着互联网的兴起和发展，现有的信任评估机制还存在一些不足，如不能快速有效地处理针对评估机制的恶意攻击，不能满足现实节点对评估准确性的要求，不能有效地识别和预防共谋节点的攻击等。针对以上不足，本文对信任评估机制做了深入研究，并提出了一种基于改进D-S证据理论的可信评估机制。所做出的具体工作如下：首先，提出了一种基于改进D-S证据理论的信任评估模型及其实现框架。结合D-S证据理论对信任的表达方式用一个三元组重新给出了定义，从而考虑了信任的不确定特性；改进基本可信度函数，用持续序列因子来解决网络实体的振荡攻击和欺骗攻击行为；给出计算持续序列因子的具体算法，使模型的实用性更强；对基于D-S证据理论的三元组信任关系，，提出了基于评估函数的归一化方法，使信任的度量更准确。实验证明本模型能更快地抑制单体恶意攻击行为，并且使评估结果更贴近实际值。其次，提出了基于G-N算法的共谋节点识别模型及其部署实现。结合社会学关系对网络团体的共谋行为进行分析得出相互吹捧的行为特征，为后期的聚类做进一步的准备；给出计算相互吹捧系数的具体算法，让算法的实用性更强；设计可疑共谋节点的社区划分方法，进一步缩小备选项的数据集，使算法的效率更高；通过评估各可疑团体的共谋程度和共谋危害识别出共谋节点，使该信任模型能更好地抑制共谋攻击行为。实验证明该识别模型在识别共谋节点方面有较高的准确性，增强了信任评估机制的安全性和可靠性。最后，基于以上的理论基础，本文设计并实现了基于改进D-S的信任评估机制的仿真系统，并进行了仿真实验以及结果分析，验证了本模型具有更快地抑制单体恶意攻击行为和共谋攻击行为，提高了评估结果的准确性。
[Abstract]:In recent years, with the rise and development of the Internet, the existing trust evaluation mechanism has some shortage, such as can not be handled quickly and efficiently according to the evaluation mechanism of malicious attacks, can not meet the requirements of the evaluation accuracy of realistic node, can not effectively identify and prevent collusion nodes attack. To solve the above problems, this paper makes a deep mechanism research on trust evaluation, and proposes a trust evaluation mechanism based on improved D-S evidence theory. The specific work made as follows:
First of all, proposed a trust evaluation model based on improved D-S evidence theory and its implementation framework. Combined with the expression of D-S evidence theory of trust with a three tuple redefined, and considering the uncertain characteristics of trust; improve the basic probability function, using continuous sequence factor to resolve network entity attacks and oscillation spoofing attack behavior; calculating specific algorithm for series factor, the more practical model of trust; three tuple based on D-S evidence theory, proposed normalization method based on evaluating function, the trust metric is more accurate. The experiment proved that the model can quickly suppress single malicious attacks, and make the assessment it is more close to the actual value.
Secondly, put forward to realize common node recognition model of G-N algorithm and its deployment. Based on the combination of Sociology of network group collusion analyze the behavior characteristics of praise from each other to make further preparations for later clustering; arithmetic backslapping coefficients are given, so that more practical algorithm; community division method suspected collusion nodes, further narrowing the options data set, make the algorithm more efficient; by assessing various suspicious groups collusion and collusion harm degree identify collusion nodes, so that the trust model can effectively reduce the collusion behavior. Experimental results show that the recognition model has higher accuracy in recognition of common nodes. To enhance the safety and reliability of the trust evaluation mechanism.
Finally, based on the above theory, this paper designed and implemented the simulation system of trust evaluation mechanism based on improved D-S, and the simulation and analysis of the results, proved this model has faster inhibition of monomer malicious attacks and collusion attack behavior, improve the accuracy of the evaluation results.

【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.08

【参考文献】