基于用户行为的网络病毒传播模型研究

Our society is confronting more threats of computer viruses and worms with the rapid growth of intelligent terminal devices. The propagation and diffusion processes of these viruses and worms are very complicated, which are closely related to human behaviors. Due to the destructiveness, researchers cannot study the characteristics of virus propagation in the real world. Propagation models, as reliable test-beds, are used to uncover the underlying characteristics of virus diffusion, gain a deep understanding of transmission mechanisms and nonlinear relationships behind the propagation process. However, current propagation models focus mainly on the effects of propagation threshold and network topology on virus propagation, which ignore/simplify the impact of the diversity and dynamics of human behaviors. In this regard, our project emphasizes and unveils the impact of human behaviors on virus propagation based on the theories of human dynamics and social psychology. By means of analyzing the temporal-spatial characteristics of empirical data, collected from the log files of emails and mobile phones, we aim to confirm several key parameters that affect human behaviors, so that a specific behavior-based propagation model can be built. First, this project will qualitatively and quantitatively analyze human operational behaviors and mobile behaviors. Utilizing the group analysis methods (e.g., the statistical method of Structural Equation Modeling (SEM)), we can construct different behaviors dynamic models through distinguishing different types of users’ behaviors in order to realize the diversity of human behaviors in the propagation model. Then, this project will analyze the process of human decision and quantify several impact factors that affect the changes of human behaviors from the viewpoint of social psychology, in order to realize the dynamic of human behaviors in the propagation model. Based on the above two kinds of behavior analysis, we can enhance and improve existing virus propagation models through characterizing the different types of users’ behaviors and their decision making process. The significance of this project lies in that it will reveal the underlying temporal-spatial characteristics of human behaviors during virus propagation and provide theoretical foundations for restraining virus propagation in the real world.

随着各类智能终端设备的普及，以及电子邮件、社交网络的普遍使用，依附于这些网络上的病毒给社会构成了巨大威胁。由于这些病毒的传播受多种人为因素影响，传播过程极其复杂。鉴于病毒危害巨大，无法在真实环境中研究其传播规律，各类传播模型为揭示病毒传播共性、挖掘隐藏在传播表象之下的各种传播机理和非线性关系做出了重要贡献。然而现有模型大都以传播阈值和网络拓扑结构对传播的影响为主要研究内容，忽略了用户行为多样性与动态性对传播的影响。本项目从邮件和手机用户行为数据的统计分析入手，利用结构方程模型区分用户行为差异，构建不同类型用户行为动力学方程，实现对用户行为多样性的刻画；利用行为改变理论，确定用户行为改变的关键因素及量化方法，刻画病毒传播过程中的用户决策过程，实现对用户行为动态性的刻画。在此基础上，构建基于用户行为的病毒传播模型, 揭示用户行为的时空特征对病毒传播的影响，为抑制网络病毒的传播扩散提供理论依据。

当今诸多现实传播现象，如传染病蔓延、计算机病毒传播和社交网络谣言扩散等，都可建模为某一传播现象在特定传播网络上的传播动力学过程。复杂网络传播动力学研究方法对揭示传播共性规律、预测传播趋势、控制传播过程和挖掘隐藏在传播表象之下的各种传播机理和非线性关系有着重要理论意义和应用前景。然而，现有传播模型都以传播阈值和网络拓扑对传播的影响为主要研究内容，忽略了个体行为多样性和动态性对传播的影响。本项目从用户行为数据时空特征分析、基于多Agent系统的行为建模和网络动力学理论及其应用三个方面展开研究。首先，从实证数据出发，分类挖掘和表示个体行为特征，并利用多Agent系统实现传播过程中个体行为决策建模，在基于用户行为的传播模型中实现对用户行为多样性和动态性的刻画。其次，利用多Agent系统揭示一种单细胞生物觅食过程中展现的自组织计算特性，并将其用于诸如网络路径规划、社团挖掘、传播源定位等多种难计算问题求解中。最后，项目组进一步分析了网络结构特征对传播的影响，揭示节点中心性对传播过程的相对促进和抑制作用。. 在该项目的资助下，项目累计发表标注资助论文31篇，其中SCI收录16篇，EI检索会议13篇：包括CCF-C类期刊4篇，CCF-C类会议2篇；国内一级学报1篇(EI)；中科院JCR二区论文5篇，三区论文7篇；被国际知名杂志或会议多次引用和评价，google引用56次。