基于机器学习的Web用户行为认证  

Behavior authentication of Web users based on machine learning

作　　者：毋泽南 田立勤 王志刚[2] 

机构地区：[1]华北科技学院计算机学院,北京101601 [2]青海师范大学计算机学院,青海西宁810008

出　　处：《网络与信息安全学报》2018年第1期45-51,共7页Chinese Journal of Network and Information Security

基　　金：国家自然科学基金资助项目(No.61472137);中央高校基本科研业务费基金资助项目(No.3142015022;No.3142017053);青海省重点研发;应用基础研究基金资助项目(No.2016-SF-130;No.2017-ZJ-752);河北省物联网监控工程技术研究中心基金资助项目(No.3142016020);青海省物联网重点实验室基金资助项目(No.2017-ZJ-Y21)~~

摘　　要：针对Web用户信息的安全问题,结合机器学习的方法,对用户行为进行分析和认证。首先通过主成分分析法对原始数据集做降维处理,然后利用SVM算法,让计算机对历史用户行为证据进行学习,得到一个判别用户身份的模型。实际应用和理论分析表明,该模型在用户行为认证判别上,可以更准确和高效地分类出危险用户和可信用户,为诸如电子商务、网络金融等关键网络应用用户行为的高性能分析奠定坚实的理论和实践基础。According to the security problem of Web user information, the user behavior was analyzed and authenticated by the method of machine learning. First of all, through the principal component analysis to reduce the dimen- sion of the original data set, then use the SVM algorithm to allow the computer to learn the history of user behavior evidence, to get a model to identify the user＇s identity. The practical application and theoretical analysis show that the model in user behavior identification authentication, can be more accurate and efficient classification of danger- ous users and trusted users, analysis lay a solid theoretical and practical basis for the high performance user behavior such as electronic commerce, network finance and other key of Interact applications.

关 键 词：信息安全 用户行为认证 支持向量机 主成分分析 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]