摘要： 网络入侵检测所处理的数据由多类攻击数据和正常数据构成，基于此对多分类支持向量机在网络入侵检测中的应用进行了研究，采用一对一方法构造了多分类支持向量机分类器，用KDD99入侵检测数据对所提出的多分类支持向量机分类器进行了测试评估，将实验结果和BP神经网络方法进行了比较。实验表明提出的方法是可行的、高效的。

关键词: 入侵检测, 统计学习理论, 多分类支持向量机, 核函数

Abstract: Network intrusion detection data are made up of multi-class attacks data and normal data. The application of multi-class support vector machine(SVM) for network intrusion detection is researched. The multi-class support vector machine is designed to detect network intrusion by using one-against-one method. The designed SVM classifier is evaluated with KDD99 intrusion detection dataset, the results obtained are compared with BP neural networks. Experimental results show that classifier based on multi-class SVM is effective and efficient.

Key words: intrusion detection, statistical learning theory(SLT), multi-class SVM, kernel function

中图分类号: 

• TP393.08