基于互信息特征选择和LSSVM的网络入侵检测系统  

Network intrusion detection system based on mutual information feature selection and LSSVM

作　　者：庄夏[1] 

机构地区：[1]中国民航飞行学院科研处,四川广汉618307

出　　处：《中国测试》2017年第11期134-139,共6页China Measurement & Test

基　　金：国家自然科学基金民航联合基金重点项目(U1233202/F01)

摘　　要：为提高网络入侵检测系统(IDS)的性能,提出一种基于互信息特征选择和LSSVM的入侵检测方案(BMIFSLSSVM)。将采集到的网络连接数据进行规范化处理,并提出一种权衡考虑特征相关性和冗余性的新型互信息特征选择(BMIFS)方法,以此从网络连接数据中选择出有效特征集。根据提取的训练样本特征集,利用最小二乘支持向量机(LSSVM)构建分类器和简化粒子群优化(SPSO)算法来优化LSSVM的核函数宽度系数和正则化参数,最后利用训练好的分类器进行入侵检测。仿真结果表明:提出的BMIFS能够选择出最优特征集,使BMIFS-LSSVM提高入侵检测准确率且降低误报率。In order to improve the performance of network intrusion detection system(IDS),an intrusion detection scheme based on mutual information feature selection and LSSVM was proposed(BMIFS-LSSVM).First,the collected network connection data was normalized.Then,a new mutual information feature selection(BMIFS) method was proposed to balance the feature correlation and redundancy,and a effective feature set was selected from the network connection data.Then,the least squares support vector machine(LSSVM) was used to construct the classifier according to the extracted training sample feature set,and the simplified particle swarm optimization(SPSO) algorithm was used to optimize the kernel function width and regularization parameters of LSSVM.Finally,the trained classifier was used for intrusion detection.The simulation results show that the proposed BMIFS can select the optimal feature set,make the BMIFS-LSSVM improve the intrusion detection accuracy and reduce the false alarm rate.

关 键 词：网络入侵检测 互信息特征选择 最小二乘支持向量机 简化粒子群优化 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]