基于相似属性PCA和SVM的网络入侵检测

针对现有网络入侵检测算法泛化能力差与处理大样本数据耗时长的问题,本文提出了基于相似属性主成分分析（PCA）与支持向量机（SVM）的网络入侵检测的方法。采用KDD 1999数据集仿真,原始数据集根据属性间的相似程度分为四类属性集,对四类属性集分别采用PCA进行特征抽取,最后用SVM检验分类的正确率。实验结果表明：与直接采用PCA对全部属性一起降维相比,相似属性PCA的分类降维方法有较短的处理时间,并且有更强的泛化能力,即对未知攻击类型的检测性能。

Network Intrusion Detection Based on PCA for Similar Attributes and SVM

In view of the poor generalization ability and long time in processing large sample data ot existing network intrusion detection algorithm, This paper proposes a intrusion detection method based on principal component analysis （PCA） for similar attributes and support vector machine （SVM）. The KDD 1999 dataset is used to achive the experiment, first, the original dataset is divided into four types of attribute sets according to the similarity between the attributes, then we use PCA to reduce the dimensions for four types of attribute sets, respectively. The result shows that comparing to the result to reduce the dimensions for all attributes, this method not only has shorter processing time, but also has stronger generalization ability, i.e. Better performence for the unknown attrack types.