特征和分类器参数组合优化的网络入侵检测.pdf

第41卷 第1期 南京理工大学学报 Vol.41 No.1 2017年2月 Journal of Nanjing University of Science and Technology Feb.2017 特征和分类器参数组合优化的网络入侵检测 王战红 (铁道警察学院公安技术系,河南 郑州450053) 摘摇 要:为了提高网络入侵检测的入侵检测结果,该文设计了特征和分类器参数组合优化的网 络入侵检测算法。 分别分析了特征、分类器参数对入侵检测结果的影响,并建立了两者组合优 化的数学模型,采用生物地理学优化算法模拟生物种群聚居栖息地的迁移过程对数学模型的最 优解进行优化,找到最优的特征和分类器参数组合,最后采用标准入侵检测数据集—KDD Cup 99 对算法的可行性和优越性进行测试和分析。 结果表明,该文算法充分利用了特征和分类器 参数之间的关联,改善了入侵检测率,执行速度可以满足入侵检测的实时性要求。 关键词:网络入侵;特征选择;分类器设计;生物地理学优化算法 中图分类号:TP393.08摇 摇 文章编号:1005-9830(2017)01-0059-08 DOI:10.14177/ ki.32-1397n.2017.41.01.008 Network intrusion detection by using combination optimizing features and classifier parameters Wang Zhanhong (Department of Police Technology,Railway Police College,Zhengzhou450053,China) Abstract:In order to obtain better intrusion detection results,this paper designs a network intrusion detection algorithm by using combination optimizing features and classifier parameters.A mathematical model of combinatorial optimization is set upbasedonthefeaturesandparametersof classifier influence on intrusion detection results respectively.A biogeography鄄based optimization algorithm is adopted to simulate migrationprocessof speciesinhabitancy tofindthe optimal solutionof mathematical model and obtainthe optimal featuresand classifier parameters.Standardintrusion detection-KDDCup99 data sets are usedto test feasibility and superiority.The results show that the proposed algorithm can make mine relation between features and classifier parameters to improve intrusion detection rate and that the execution speed can meet the real-time requirementsof intrusion detection. Key words:network intrusion;feature selection;classifier