面向网络公害社群的社交网络特性分析方法

Analysis method on the properties of social networks of internet nuisance communities

随着网络的普及和网民群体的壮大,网络水军、网络公害社群等互联网公害组织严重威胁我国网络环境的安全。针对这类组织远离国内监管、隐蔽性和伪装性强的特点,提出了面向网络公害社群的社交网络特性分析方法。在用户特性层面,通过中心性测量对组内成员的影响力分布进行计算,找出影响力最大的核心用户,并分析核心成员的匿名性特点。在复杂网络特性层面,通过小世界效应和无标度特性的分析,研究了该社群内部的集群状况和资源分布状况。在时序特性层面,对核心成员的发文内容进行预处理后,利用BERTopic进行动态主题建模,得到社群主题随时间的变化关系。大量的实验性能分析验证了本文方法的有效性,研究成果可用于分析公害社群相关特征,追踪犯罪活动轨迹,为公害组织的侦查和打击工作以及自动化检测系统的设计工作提供理论参考。

With the widespread adoption of the Internet and an increase in the number of Internet users,the security of the network environment has been seriously threatened by Internet public hazards organizations,such as the In‐ternet water army and the internet public hazards community.Given the characteristics of these organizations,which include being distant from domestic supervision and possessing strong concealment and camouflage abili‐ties,a social network characteristic analysis method was proposed for Internet public hazards organizations.Ini‐tially,at the user characteristic level,the influence distribution of group members was calculated through centrality measurement to identify the core users with the greatest influence,and an analysis of the anonymity characteristics of these core members was conducted.Subsequently,at the complex network characteristic level,the cluster status and resource distribution within the community were studied through the analysis of small-world effects and scale-free characteristics.Finally,at the temporal characteristic level,after preprocessing the post content of core mem‐bers,BERTopic was utilized for dynamic topic modeling to ascertain the relationship between community themes and time.A large number of experimental performance analyses have verified the effectiveness of the proposed method.The research results can be utilized to analyze the characteristics of public hazard communities,track criminal activity trajectories,and provide theoretical references for the investigation and crackdown of public haz‐ard organizations,as well as for the design of automated detection systems.