在线社交网络中异常帐号检测方法研究

Anomaly Detection in Online Social Networks

异常帐号检测是在线社交网络安全研究的关键问题之一.攻击者通过异常帐号传播广告、钓鱼等恶意消息以及恶意关注、点赞等行为严重威胁正常用户的信息安全和社交网络的信用体系,为此有大量的研究工作来检测社交网络中异常帐号.文中回顾了近年来在线社交网络中异常帐号检测的主要成果,阐述了异常帐号在不同发展阶段的表现形式以及检测异常帐号所面临的主要挑战,重点从基于行为特征、基于内容、基于图(Graph)、无监督学习四个方面总结了异常帐号检测方案,介绍了在实验过程中数据获取、数据标识以及结果验证的主要方法,并对未来异常帐号检测的研究趋势进行了展望.

Anomaly detection is increasingly becoming a focus in the field of Online Social Networks（OSNs）security.Attackers have utilized OSNs as new platforms to conduct malicious behaviors,which have seriously threatened the privacy of users and the reputation of OSNs,including sending spam,phishing and other illicit activities such as selling followers and Page Like.Many detection techniques have been specifically developed in past years for spotting anomalies in OSNs.This paper reviews important achievements in anomaly detection in recent years.First,different behaviors of anomalies are elaborated with the grand challenges to the detection.Second,we discuss the detection techniques with respect to feature-based,contentbased,graph-based and unsupervised approaches.Third,major methods of collecting datasets,labeling anomalies and validating results are introduced.Finally,we conclude the paper with an exploration of future research directions on anomaly detection in OSNs.