一种适于SVM入侵检测的余弦时变粒子群方法  

A Cosine Time-varying Particle Swarm Optimization for SVM Intrusion Detection

作　　者：杨红浩 周治平[1,2] YANG Hong-hao;ZHOU Zhi-ping(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China;Engineering Research Center of lnternet of Things Technology Applications Ministry of Education,Jiangnan University,Wuxi 214122,China)

机构地区：[1]江南大学物联网工程学院,江苏无锡214122 [2]江南大学物联网技术应用教育部工程研究中心,江苏无锡214122

出　　处：《控制工程》2020年第9期1589-1594,共6页Control Engineering of China

基　　金：江苏省研究生科研与创新计划项目(SJCX17_0507)。

摘　　要：针对时变粒子群算法对支持向量机惩罚参数和核函数参数寻优能力不足导致检测效率相对较低的问题,提出一种余弦时变粒子群优化支持向量机的方法。该方法通过运用余弦时变加速系数平衡算法在全局和局部的搜索能力,根据粒子适应度值与自定义的平均适应度值之间的距离,修改惯性权重的取值策略。结合2种方法来优化粒子群算法,控制粒子全局搜索和快速收敛到全局最优解;分别从NSL-KDD数据集和ADFA-LD数据集选取样本进行测试,仿真实验结果表明,改进后的粒子群算法能够在搜索过程中有效地避免陷入局部最优的情况,可以更加快速地搜寻确定支持向量机模型参数的最优值,提高支持向量机的分类准确率。The problem of low detection efficiency,which caused by lack of search optimal ability,which of the penalty parameters and kernel function of the Support Vector Machine(SVM),the method based on SVM,which optimized by Cosine Time Varying Particle Swarm Optimization(CTVPSO)is proposed.It uses the cosine time varying acceleration coefficients,to balance the exploration and exploitation abilities,according to the distance between the fitness value and the custom average fitness value,to modify the setting strategy of inertia weight,then combine two methods to improve the Particle Swarm Optimization(PSO),to control particle search in the global scope,and converge to the global optimal solution quickly.Selected sample data from NSL-KDD datasets and ADFA-LD datasets for testing,respectively,the experiments demonstrate that this method can improve the performance of the PSO algorithm remarkably,avoiding it falling into local optimum,and faster in searching the optimum,to improve the classification accuracy of the SVM.

关 键 词：入侵检测 余弦时变 粒子群算法 支持向量机 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]