基于自适应蛙跳算法的入侵检测特征选择  

Intrusion detection feature selection based on adaptive frog leaping algorithm

作　　者：彭平[1] 孙立新 王铁柱[3] 李恒建[4] 

机构地区：[1]河北中医学院公共课教学部,河北石家庄050001 [2]河北省疾控中心药械处,河北石家庄050001 [3]河北中医学院信息技术中心,河北石家庄050001 [4]济南大学信息科学与工程学院,山东济南250022

出　　处：《济南大学学报（自然科学版）》2016年第2期129-132,共4页Journal of University of Jinan(Science and Technology)

基　　金：国家自然科学基金(61303199)

摘　　要：针对目前蛙跳算法应用到入侵检测特征选择时均存在容易陷入局部最优、迭代后期收敛速度慢等问题,提出一种基于自适应蛙跳算法的特征选择方法。该方法采用自适应的变异策略,通过层次分析法自适应调整各影响参数权重以及建立模糊判断矩阵选择调整参数概率来改进蛙跳算法。实验结果表明:改进的算法显著提高了收敛性能,具有很强的自适应能力,不但能对不同类型攻击进行检测,并且对不同类型攻击具有较好的均衡性;在保证检测率较高的同时也具有较低的误报率。In the application of intrusion detection feature selection,the current frog leaping algorithm is easy to fall into local optimum and slowing convergence speed in the late iteration,so an intrusion detection feature selection method based on the adaptive frog leaping algorithm is presented. Adopting the adaptive mutation strategy,this method adaptively adjusts parameter weights through analytic hierarchy process and establishes the fuzzy judgment matrix to choose the probabilities parameters. Experimental results prove that the improved adaptive frog leaping algorithm improves the convergence performance significantly and has strong adaptive ability,with a higher detection rate and a low rate of false positives on the basis of the different types of attack detection.

关 键 词：蛙跳算法 智能算法 仿生算法 特征选择 入侵检测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]