基于改进降噪自编码模型的工业控制系统入侵识别研究  

Research on Intrusion Identification of Industrial Control Systems Based on Improved Denoising Autoencoder Model

作　　者：寇宏波 KOU Hongbo(Everbright Environmental Protection(China)Co.,Ltd.,Shenzhen,Guangdong 518000,China)

机构地区：[1]光大环保(中国)有限公司,广东深圳518000

出　　处：《自动化应用》2024年第17期69-72,共4页Automation Application

摘　　要：传统的工业控制系统入侵识别方法在面对复杂的攻击模式和隐蔽的攻击手段时,往往存在漏报、误报等问题,为此,提出基于改进降噪自编码模型的工业控制系统入侵识别方法。首先,处理与提取工业控制系统数据特征,进行分箱操作,构建改进降噪自编码入侵识别模型。然后,在隐层节点后添加Softmax函数分类器进行分类,完成智能识别工业控制系统入侵的操作。结果表明,所设计方法能够准确识别测试数据集中各类入侵行为数量,且识别准确率较高,可实现对工业控制系统的有效防护。Traditional intrusion recognition methods for industrial control systems often suffer from issues such as missed and false alarms when facing complex attack patterns and covert attack methods.Therefore,an improved denoising autoencoder model based intrusion recognition method for industrial control systems is proposed.Firstly,process and extract data features from industrial control systems,perform binning operations,and construct an improved denoising autoencoder intrusion recognition model.Then,a Softmax function classifier is added after the hidden layer node for classification,completing the operation of intelligent recognition of industrial control system intrusion.The results show that the designed method can accurately identify the number of various types of intrusion behaviors in the test dataset,with high recognition accuracy,and can achieve effective protection for industrial control systems.

关 键 词：入侵识别方法 工业控制系统 模型训练 改进降噪自编码模型 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]