基于决策树算法的网络入侵检测系统设计与评估  

Design and evaluation of network intrusion detection system based on decision tree algorithm

作　　者：张小云 康晓霞 ZHANG Xiao-yun;KANG Xiao-xia(Gansu Corps of Chinese People’s Armed Police Force,Lanzhou 730046,China)

机构地区：[1]中国人民武装警察部队甘肃总队,兰州730046

出　　处：《信息技术》2023年第2期117-122,共6页Information Technology

摘　　要：实时网络数据包含大量冗余术语和噪声,而现有入侵检测技术准确度较低,特征提取能力不足。针对NSL-KDD数据集,提出了一种基于决策树的网络入侵检测系统。采用相关特征选择子集评价方法进行特征选择并减小维数,消除冗余数据,提高资源利用率并降低时间复杂度,通过特征选择可提高入侵检测方法预测性能。在特征选择之前和特征选择之后,对五类分类和二类分类进行性能评估。结果表明,该系统具有较高检出率和精度,数据集二类分类总体结果高于五类分类,可为网络安全检测工作提供借鉴。Real time network data contains a large number of redundant terms and noise,while the existing intrusion detection technology has low accuracy and insufficient feature extraction ability.Based on NSL-KDD data set,a network intrusion detection system based on decision tree is proposed.Correlation feature selection(CFS)subset evaluation method is used to select features and reduce dimension,eliminate redundant data,improve resource utilization and reduce time complexity.Feature selection can improve the prediction performance of intrusion detection methods.Before and after feature selection,the performance of class V and class II classification is evaluated.The results show that the system has high detection rate and accuracy,and the overall result of class II classification is higher than that of class V classification.It can provide reference for the work of network security detection.

关 键 词：决策树 特征选择 入侵检测系统 NSL KDD数据集 相关特征选择 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]