基于互信息与萤火虫算法的网络入侵特征选择  

NETWORK INTRUSION FEATURE SELECTION BASED ON MUTUAL INFORMATION AND FIREFLY ALGORITHM

作　　者：王新胜[1] 杨锐 Wang Xinsheng;Yang Rui(School of Computer Science and Communication Engineering,Jiangsu University,Zhenjiang 212013,Jiangsu,China)

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013

出　　处：《计算机应用与软件》2024年第4期306-312,320,共8页Computer Applications and Software

基　　金：国家自然科学基金项目(61402205);江苏省自然科学基金青年项目(BK20190838)。

摘　　要：为减少网络入侵检测数据中的冗余特征,提出一种结合互信息和萤火虫算法的特征选择方法。针对互信息不能精确计算特征间冗余度,提出类内特征冗余互信息特征选择方法。针对萤火虫算法步长因子固定易使算法陷入局部最优等问题,提出自适应步长萤火虫算法特征选择。以上方法分别选取特征子集后利用投票策略选取最优子集,对该子集基于C4.5和贝叶斯网络分类器分类。实验结果表明,使用10个特征检测能有效提高入侵检测率、误报率和F-measure,同时还缩短训练和检测时间。此外,与现有的几种方法相比,该方法在准确率、检测率和F-measure都获得不错效果。In order to reduce redundant features in network intrusion detection data,this paper proposes a feature selection method based on mutual information and firefly algorithm.Aimed at the imprecision calculation of redundancy between features for mutual information,a feature selection method for inner class feature redundancy mutual information was proposed.In order to solve the problem that the fixed step factor in firefly algorithm made the algorithm fall into local optimum,the feature selection of adaptive step size firefly algorithm was proposed.After the feature subset was selected by the above methods,the optimal subset was selected by using the voting strategy.The intrusion detection based on C4.5 and Bayesian network classifier was carried out for this subset.The experimental results show that using 10 features can effectively improve the intrusion detection rate,false alarm rate and F-Measure,and also shorten the training and detection time.In addition,compared with the existing methods,this method achieves good results in accuracy,detection rate and F-Measure.

关 键 词：网络入侵检测 特征选择 投票策略 互信息 萤火虫算法 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]