计算机工程与应用 ›› 2012, Vol. 48 ›› Issue (5): 1-3.

• 博士论坛 • 上一篇    下一篇

采用改进PTSVM的入侵检测研究

刘 宇1,2,朱随江1,2,刘宝旭2   

  1. 1.中国科学院 研究生院,北京 100049
    2.中国科学院 高能物理研究所 计算中心,北京 100049
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2012-02-11 发布日期:2012-02-11

Study on intrusion detection using improved Progressive Transductive SVM

LIU Yu1,2, ZHU Suijiang1,2, LIU Baoxu2   

  1. 1.Graduate University, Chinese Academy of Sciences, Beijing 100049, China
    2.Computing Center of Institute of High Energy Physics, Chinese Academy of Sciences, Beijing 100049, China
  • Received:1900-01-01 Revised:1900-01-01 Online:2012-02-11 Published:2012-02-11

摘要: 针对ISVM以及TSVM在基于异常的入侵检测中存在的问题,面向网络入侵数据特征改进了PTSVM算法,提出了有倾向的区域标注法,使其可以快速准确地对以无标签训练样本为主的入侵数据进行训练学习,得到接近最优解的分类超平面。实验证明基于改进PTSVM的入侵检测算法在训练和检测速度上明显高于其他算法,对于攻击样本的检测率有很大提高。

关键词: 网络安全, 入侵检测, 半监督学习, 渐进直推式支持向量机, 有倾向区域标注

Abstract: To overcome the problems of ISVM and TSVM in intrusion detection, this paper improves PTSVM for network intrusion data features and designs a method with area marked tendency. It makes the training of PTSVM more quickly and accurately, leading a split plane more close to the optimum. Experiments show that intrusion detection based on improved PTSVM has an obvious advantage in training and testing speed, and it greatly improves the detection rate of attack samples.

Key words: network security, intrusion detection, semi-supervised learning, progressive transductive Support Vector Machine(SVM), area marked tendency