恶意社交机器人检测技术研究

发布时间：2018-10-21 09:38

【摘要】：攻击者利用恶意社交机器人窃取用户隐私、传播虚假消息、影响社会舆论,严重威胁了个人信息安全、社会公共安全,乃至国家安全。攻击者还在不断引入新技术实施反检测。恶意社交机器人检测成为在线社交网络安全研究的一个重点和难点。首先回顾了当前社交机器人的开发与应用现状,接着对恶意社交机器人检测问题进行了形式化定义,并分析了检测恶意社交机器人所面临的主要挑战。针对检测特征的选取问题,厘清了从静态用户特征、动态传播特征,以及关系演化特征的研究发展思路。针对检测方法问题,从基于特征、机器学习、图论以及众包4个类别总结了已有检测方案的研究思路,并剖析了几类方法在检测准确率、计算代价等方面的局限性。最后,提出了一种基于并行优化机器学习方法的恶意社交机器人检测框架。
[Abstract]:Attackers use malicious social robot to steal user privacy, spread false information, influence public opinion, and threaten personal information security, social public safety, and even national security. Attackers are also constantly introducing new technologies to carry out counter-detection. Malicious social robot detection has become an important and difficult point in online social network security research. Firstly, the development and application of social robot are reviewed, then the problem of malicious social robot detection is formally defined, and the main challenge to detect malicious social robot is analyzed. Aiming at the selection of detection features, this paper clarifies the research and development of static user features, dynamic propagation features, and relationship evolution features. Aiming at the problem of detection methods, this paper summarizes the research ideas of existing detection schemes from four categories: feature based, machine learning, graph theory and crowdsourcing, and analyzes the limitations of several kinds of methods in detection accuracy and computation cost. Finally, a detection framework for malicious social robots based on parallel optimized machine learning is proposed.
【作者单位】：南京师范大学计算机科学与技术学院;江苏省大规模复杂系统数值模拟重点实验室;
【基金】：赛尔网络下一代互联网技术创新基金资助项目(No.NGII20160509) 江苏省高等教育教学改革重点课题基金资助项目(No.2015JSJG034)~~
【分类号】：TP242;TP393.08

【相似文献】