基于改进蚁群求解特征子集的入侵检测方法

INTRUSION DETECTION METHOD BASED ON IMPROVED ANT COLONY OPTIMIZATION TO SOLVE THE FEATURE SUBNET

为了去除冗余特征,提高入侵检测系统的检测性能,提出一种基于改进蚁群算法求解特征子集的入侵检测方法。对蚂蚁的初始位置、启发函数、信息素更新策略及状态转移概率函数均做了优化。对KDD CUP 99数据集进行预处理,根据信息熵理论对特征进行初步提取。由提取的特征点构造邻接拓扑,用改进蚁群算法进一步求解特征子集。通过十折交叉验证法训练及优化改进蚁群算法及支持向量机的参数,并测试该入侵检测方法的性能。实验结果表明,相比其他方法,所提方法的性能较优,F-Measure值有一定提升,测试时间显著减少。

In order to reduce the redundant features and enhance the performance of the detection of intrusion detection system(IDS),the intrusion detection method based on improved ant colony optimization(ACO)to solve the feature subnet is proposed.The initial positions of ants,heuristic function,pheromone updating strategy and state transition probability function were all optimized in improved ACO.The datasets of KDD CUP 99 were preprocessed and the features were initially extracted according to the information entropy theory;the adjacent topologies were structured with extracted feature nodes and the feature subsets were further obtained by improved ant colony optimization;the parameters of improved ACO and support vector machine(SVM)were optimized by training,and the performance of the proposed intrusion detection method was testified using ten-fold cross-validation on KDD CUP 99 data set.The experimental results show that compared with other methods,the proposed intrusion detection method performs better,the F-Measure value is enhanced to some extent and the testing time is reduced significantly.