作　　者： (吴大鹏）; (司书山）; (闫俊杰）; (王汝言）;

机构地区： 重庆邮电大学通信与信息工程学院,重庆400065 重庆高校市级光通信与网络重点实验室,重庆400065

出　　处： 《电子与信息学报》 2017年第9期2089-2096,共8页

摘　　要： 通过制造大量非法虚假身份,女巫攻击者可以提高自身在社交网络中的影响力,影响网络中社交个体中继选择意愿,窃取社交个体隐私,对其利益造成严重威胁。在对女巫节点行为特征分析的基础上,该文提出一种适用于社交网络的女巫节点检测机制,通过节点间静态相似度和动态相似度评估节点影响力,并筛选可疑节点,进而观察可疑节点的异常行为,利用隐形马尔科夫模型推测女巫节点通过伪装所隐藏的真实身份,更加精确地检测女巫节点。分析结果表明,所提机制能有效提高女巫节点的识别率,降低误检率,更好地保护社交个体的隐私和利益。 Sybil attackers can improve their own influence in social networks by creating a large number of illegal illusive identities then affect the social individuals＇ choice of relays and steal individuals＇ privacy, which seriously threatens the interests of social individuals. Based on the analysis of the Sybil＇s behaviors, a Sybil detection mechanism applied to social networks is proposed in this paper. The influence of nodes is calculated according to static similarity and dynamic similarity and then selecting the suspicious nodes based on the influence. Next, using the Hidden Markov Model （HMM） to infer the true identity of suspicious nodes by observing their abnormal behaviors, thus detecting the Sybil more precisely. Analysis results show that the proposed mechanism can ef- fectively improve the recognition rate and reduce the false detection rate of the Sybil and thereby protecting the privacy and interests of social individuals better.

关 键 词： 社交网络 女巫节点检测 行为特征 隐形马尔科夫模型