社交网络中基于随机游走介数的Sybil攻击检测算法研究

发布时间：2018-04-23 15:38

本文选题：社交网络 + 攻击检测　；参考：《燕山大学》2014年硕士论文

【摘要】：随着社交网络的迅猛发展，越来越多的用户通过社交网络沟通交流、分享信息，然而由于社交网络的开放性，社交网络用户更容易受到安全威胁，尤其是Sybil攻击呈上涨趋势，一些恶意用户为了谋求利益，创建大量的恶意身份，向社交网络中真实用户传播恶意信息或者提升自己团体的影响，严重威胁着社交网络的安全。本文在综合分析国内外研究现状的基础上，针对如何在解决大规模社交网络中Sybil攻击检测问题进行了深入地研究。首先，针对现有Sybil攻击检测算法假设相对严格并且计算代价高，不能有效应用在较大规模的社交网络的问题，通过分析Sybil攻击模型的特点，攻击节点需要经过攻击边对系统实施攻击，使得攻击边的边介数明显高于正常边的边介数值，提出一种更加接近于真实社交网络中信息传播的c-path边介数模型，限制随机游走路径的长度，合理选择路径出发点和游走策略，降低计算复杂度，提出使得边介数性质可以应用在较大规模的社交网络的边介数计算算法。其次，针对现有攻击检测算法没有有效检测恶意用户团体方案的问题，提出一种基于聚类的Sybil团体检测算法。该算法使用边介数结合边聚类系数作为特征，通过k-means算法进行聚类，利用种子集中的真实用户的数目确定真实边和Sybi攻击边的类簇。然后由检测得到的Sybil节点通过标签传播算法检测Sybil节点所在的恶意团体。最后，在不同的数据集上，，将本文提出的Sybil攻击检测方法和现有的检测方法进行实验对比并进行分析。
[Abstract]:With the rapid development of social networks, more and more users communicate and share information through social networks. However, because of the openness of social networks, social network users are more vulnerable to security threats, especially Sybil attacks are on the rise. In order to seek benefits, some malicious users create a large number of malicious identities, spread malicious information to real users in social networks or enhance the influence of their own groups, which seriously threaten the security of social networks. Based on the comprehensive analysis of the current research situation at home and abroad, this paper makes a thorough study on how to solve the problem of Sybil attack detection in large-scale social networks. First of all, aiming at the problem that the existing Sybil attack detection algorithms are relatively strict and computationally expensive, and can not be effectively applied to large scale social networks, the characteristics of the Sybil attack model are analyzed. The attack node needs to attack the system through attacking side, which makes the edge medium number of the attack side obviously higher than the normal edge medium value. This paper proposes a c-path edge medium model which is closer to the information propagation in the real social network. This paper limits the length of random walk path, reasonably selects the starting point and walk strategy of the path, reduces the computational complexity, and proposes an edge-medium algorithm that can be applied to large scale social networks. Secondly, aiming at the problem that the existing attack detection algorithms are not effective in detecting malicious user groups, a clustering based Sybil group detection algorithm is proposed. The algorithm uses edge mediums and edge clustering coefficients as the feature, and uses the k-means algorithm to cluster the real users in the seed set to determine the clusters of real edges and Sybi attack edges. Then the detected Sybil node detects the malicious group of Sybil nodes by tag propagation algorithm. Finally, on different data sets, the Sybil attack detection method proposed in this paper and the existing detection methods are compared and analyzed experimentally.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.08

【参考文献】