Abstract: Transductive support vector machines (TSVM) classifies the new data vector based on the information only related to this data vector. This paper proposes an anomaly network traffic detection method based on TSVM. The KDD-99 intrusion detection competition data set is used to illustrate the performances of the TSVM. The performance-comparisons of TSVM and traditional inductive SVM are presented. The results show that TSVM can be used in intrusion detection practically.

Key words: Intrusion detection, Statistical learning, Transductive support vector machine

摘要： 直推式支持向量机(TSVM) 是一种直接从已知样本出发对特定的未知样本进行识别和分类的技术。该文提出了基于TSVM的网络入侵检测系统模型，并用实验给出了它在网络入侵检测中的性能表现，分析了它与基于传统归纳式支持向量机(ISVM)的入侵检测系统的性能对比。实验结果表明，将TSVM应用到入侵检测是切实可行的。

关键词: 入侵检测, 统计学习, 直推式支持向量机