移动网络行为的多态聚类及其演化研究

项目名称： 移动网络行为的多态聚类及其演化研究

项目编号： No.61272405

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 付才

作者单位： 华中科技大学

项目金额： 81万元

中文摘要： 移动网络行为分析可挖掘出大规模数据中的内在规律，对了解用户社会特性、优化网络设计、提高服务管理水平和增强网络安全具有重要意义，是目前国内外学术和工业界共同关注的前沿科学问题。现有网络行为分析方法缺乏对时序特性的建模，不适用于多属性的研究对象，无法应对大规模移动网络场景。本课题提出多态聚类方案应对移动网络应用中需求的多样化，并采用演化聚类方法研究行为的演化特性。课题首先采用多维属性矩阵和扁平化属性链路对网络行为进行细粒度表示，结合投影寻踪理论，采用最佳投影方向和投影指标规避维数灾难问题，合理实现需求多态与行为多态的精确对应；接着从局部演化与全局演化两个角度研究网络行为的全面特征；最后以移动网络蠕虫传播与封堵作为应用案例，利用多态聚类与演化聚类方法，设计适用于移动网络的病毒防御机制。本课题的实施一方面为当前迅猛发展的移动网络提供理论与技术支撑，另一方面为移动网络的应用创新及网络安全提供新思路。

中文关键词： 移动社交网络；行为多态；演化聚类；病毒防御；网络安全

英文摘要： Network behavioral analysis is capable of discovering the inherent patterns hidden in the massive data sets, which is a cutting-edge scientific issue receiving considerable attention from academia and industry community at home and abroad, since it is bound to exert a significant effect on understanding users' social characteristics, optimizing network design, promoting network management and enhancing network security. Existing analysis techniques lack good temporal model, cannot deal with objects with multiple attributes and cannot apply to the large-scale mobile networks. In this work, we propose polymorphism clustering to meet the diverse requirements of various applications, and further we put forward the evolutionary clustering to study the temporal aspect of users' behaviors. We first express users' trace in fine grain using multi-dimensional matrix and flat property links; combining with the Project Pursuit theory to find the best projection direction and index, we effectively avoid "the curse of dimensionality" and obtain good matching between the multi-objective demand and polymorphism behavior. Then, we study the comprehensive characteristics from both the local and global evolutionary aspects. Last but not the least, we take worm propagation and blocking in mobile networks as a concrete research case

英文关键词： mobile social network；behavior Polymorphism；Evolution Clustering；Virus defense；Network Security