基于交叉验证SVM的网络入侵检测  

Intrusion Detection Based on Cross-Validation SVM

作　　者：王凯[1] 侯著荣[1] 王聪丽[1] 

机构地区：[1]军械工程学院计算机工程系,河北石家庄050003

出　　处：《测试技术学报》2010年第5期419-423,共5页Journal of Test and Measurement Technology

摘　　要：针对传统入侵检测系统漏报率和误报率高的问题,将支持向量机(SVM)应用于入侵检测中,提出了在SVM学习过程中引入交叉验证的方法,采用径向基函数(RBF)作为核,将训练集分成若干子集,每一子集使用其它子集训练得到的分类器进行测试,获得RBF的两个最佳参数后,将其应用于最终的分类器.实验结果表明,该方法能够有效检测入侵攻击,具有更高的检测率和更强的泛化能力,同时具有较低的误报率和漏报率,可以有效地运用于入侵检测系统中.Aiming at the problem of the higher rates of missing alarm and false alarm in the traditional intrusion detection system,we applied support vector machine（SVM） to intrusion detection,and proposed a novel method which introduced cross-validation in the learning procedure of SVM.Using the radial basis function（RBF） as the core,the training set was divided into several subsets,and each subset was tested by using the classifier obtained by training the other subset.The best two parameters of RBF were got and applied to the final classifier.Experimental results show that,with this method,intrusion attacks can be detected effectively and precisely,it has higher rate of detection and more generalization capability,at the same time,it has lower false alarm rate and miss rate.The method can be applied to intrusion detection system effectively.

关 键 词：机器学习 入侵检测 统计学习 支持向量机 交叉验证 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]