基于特征选择和进化神经网络的网络异常入侵检测方法  

Network anomaly intrusion detection method based on feature selection and evolutionary neural networks

作　　者：李周 马俊杰 赵灿明 杨安东 胡永杰 LI Zhou;MA Junjie;ZHAO Canming;YANG Andong;HU Yongjie(Internet Department of State Grid Anhui Electric Power Co.,Ltd.,Hefei 230022,China)

机构地区：[1]国网安徽省电力有限公司互联网部,安徽合肥230022

出　　处：《电子设计工程》2024年第17期57-60,67,共5页Electronic Design Engineering

基　　金：国网安徽省电力有限公司科技项目(B31200400006)。

摘　　要：由于网络信息数量庞大,内部存在大量冗余特征信息,异常检测时容易受其影响,导致检测效率降低,无法保障网络运行安全。为此提出基于特征选择和进化神经网络的网络异常入侵检测方法。应用主成分分析法选择合适的网络运行数据特征,基于进化神经网络构建异常入侵检测模型,阐述网络异常入侵检测过程,并制定特征数据提取模式与异常入侵判定规则,从而获取最终网络异常入侵检测结果。实验数据显示,提出方法获得异常入侵检测特征数量与最佳特征数量相同,网络异常入侵检测相对准确率最大值为98%,以此证明所提方法检测异常入侵精准性高。Due to the large amount of network information and the presence of a large amount of redundant feature information,anomaly detection is easily affected,resulting in reduced detection efficiency and inability to ensure network operation security.Therefore,a network anomaly intrusion detection method based on feature selection and evolutionary neural networks is proposed.Apply principal component analysis to select appropriate network operating data features,construct an anomaly intrusion detection model based on evolutionary neural networks,explain the process of network anomaly intrusion detection,and develop feature data extraction patterns and anomaly intrusion detection rules to obtain the final network anomaly intrusion detection results.The experimental data shows that the proposed method obtains the same number of abnormal intrusion detection features as the optimal features,and the maximum relative accuracy of network abnormal intrusion detection is 98%,which proves the high accuracy of the proposed method in detecting abnormal intrusion.

关 键 词：进化神经网络 异常入侵 特征选择 入侵检测 深度学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]