摘要
针对工业互联网攻击流量特征复杂以及深层神经网络易发生退化的问题,提出了一种基于残差网络和深度学习的入侵检测方法,实现了将一维卷积神经网络与门控循环单元残差连接的网络模型。使用CSE-CIC-IDS2018数据集和密西西比州大学的天然气管道数据集进行实验,结果表明,此方法在各个评价指标上均优于其他经典机器学习算法,具有较好检测性能和泛化能力,证明了其在工业网络环境中的可靠性及应用价值。
In view of the complex characteristics of industrial Internet attack traffic and the degradation of deep neural network,an intrusion detection method based on residual network and deep learning is proposed.The model of connecting one-dimensional convolution neural network with the residual error of gated cycle unit was realized,and the experiment was carried out using the CSE-CIC-ISD2018 data set and the natural gas pipeline data set of the University of Mississippi.The results show that this method is superior to other classical machine learning algorithms in each evaluation index,and has better detection performance and generalization ability,which proves its reliability and application value in the industrial network environment.
作者
李安娜
宗学军
何戡
连莲
Li Anna;Zong Xuejun;He Kan;Lian Lian(College of Information Engineer,Shenyang University of Chemical Technology,Shenyang 110142,China;Liaoning Provincial Key Laboratory of Information Security for Petrochemical Industry,Shenyang 110142,China)
出处
《网络安全与数据治理》
2023年第3期1-7,共7页
CYBER SECURITY AND DATA GOVERNANCE
基金
辽宁省“兴辽英才计划”(XLYC2002085)。
关键词
入侵检测
残差网络
一维卷积神经网络
门控循环单元
损失函数
intrusion detection
residual nerwork
one-dimensional convolutional neural network
gated cycle unit
loss funtion
作者简介
李安娜(1996-),女,硕士研究生,主要研究方向:工业信息安全。;通信作者:宗学军(1970-),男,硕士,教授,主要研究方向:工业信息安全。E-mail:xuejun_zong@syuct.edu.cn。;何戡(1979-),男,硕士,副教授,主要研究方向:工业信息安全。
