摘要
为了改善基于K-近邻(K-nearest neighbor,KNN)入侵检测模型的性能,提出一种基于局部搜索算法的元优化特征权重KNN入侵检测模型.利用差分进化算法优化特征权重,采用基于局部单峰采样(local unimodal sampling,LUS)的元优化模型对差分进化算法进行优化.应用NSL数据集进行仿真实验,将本优化模型和其他常用智能启发算法,包括遗传算法(genetic algorithm,GA)、粒子群优化(particle swarm optimization,PSO)算法和灰狼优化(grey wolf optimization,GWO)算法进行比较.实验结果表明,与传统KNN算法模型相比,该模型的准确率提高了2.86%,检测率提高了3.18%,误报率降低了50%,而且基于元优化的优化策略优于其他常用优化算法.
To improve the performance of intrusion detection model based on KNN,a KNN intrusion detection model using meta-optimization based on a local search algorithm for feature weighting was proposed.The differential evolution algorithm was used to optimize feature weights and the LUS based meta-optimization was selected to optimize the differential evolution.The NSL dataset was used to carry out the experiments.The proposed model was compared with that optimized by other commonly used heuristic algorithms,including GA,PSO and GWO.Results show that compared with the traditional KNN,the accuracy of the proposed method is improved by 2.86%,the detection rate increased by 3.18%and the false positive rate is reduced by 50%.The optimization based on meta-optimization is better than other optimization algorithms commonly used.
作者
沈焱萍
伍淳华
罗捷
高方平
SHEN Yanping;WU Chunhua;LUO Jie;GAO Fangping(School of Information engineering,Institute of Disaster Prevention,Beijing 101601,China;School of CyberSpace Security,Beijing University of Posts and Telecommunications,Beijing 100876,China;Patent Examination Cooperation Sichuan Center of the Patent Office,China National Intellectual Administration Property,Chengdu 610213,China)
出处
《北京工业大学学报》
CAS
CSCD
北大核心
2020年第1期24-32,共9页
Journal of Beijing University of Technology
基金
廊坊市科学技术研究与发展计划自筹经费资助项目(2017011027)
中央高校基本科研业务费青年教师资助计划资助项目(2017011027)
国家自然科学基金资助项目(61602052)
关键词
K-近邻
差分进化
元优化
局部单峰采样
机器学习
入侵检测系统
K-nearest neighbor(KNN)
differential evolution(DE)
meta-optimization
local unimodal sampling
machine learning
intrusion detection system(IDS)
