基于自适应遗传算法的入侵检测特征选择方法  

A feature selection method for intrusion detection using the adaptive genetic algorithm

作　　者：李安[1] 江加和[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《应用科技》2016年第3期49-53,59,共6页Applied Science and Technology

基　　金：国家自然科学基金项目(61273141)

摘　　要：针对网络入侵检测所处理数据特征维数高、入侵检测系统负荷大、检测速度慢等问题,提出了一种将自适应遗传算法与信息增益相结合的特征选择方法,并采用基于支持向量机的分类器作为自适应遗传算法中适应度函数的计算与特征选择结果性能的评价。实验采用入侵检测KDDCUP99数据集,将原41维特征属性约简为13维,通过和自适应遗传算法,回溯搜索算法与信息增益相结合的特征选择方法等2种算法的对比实验,表明基于自适应遗传算法的特征选择算法具有更优的解空间寻优能力和特征约简能力。An intrusion detection system （IDS） needs to deal with high-dimensional feature of data and it iseasily overloaded. Meanwhile high-dimensional feature of data will lead to low detection speed and poor detec-tion performance. To solve such defects, a novel feature selection method was proposed, combining the adap-tive genetic algorithm （GA） with information gain （ IG）, In addition, a classifier based on support vector ma-chine （SVM） is used for the calculation of the fitness function in the adaptive genetic algorithm and the per-formance assessment for the feature selection results. The experiment applied KDDCUP99 data set, and a 13-dimension feature reduction set was obtained from the original 41-dimension feature set. Compared with GAmethod and backtracking search algorithm combined with information gain （IG-BSA） method, the IG-GAmethod has better solution space optimization and feature reduction ability.

关 键 词：遗传算法 信息增益 特征选择 入侵检测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]