基于CNN-LSTM的工业控制网络入侵检测方法研究

打开文本图片集

中图分类号：TP393；TP277 文献标识码：A 文章编号：2096-4706（2025）24-0180-06

Abstract：As Industrial Control Systems （ICS）are graduall integrated into the Internet，thecybersecurity challenges theyfaceare increasinglygrowing.To address the intrusion detectionproblemin IndustrialControlNetworks（ICN），this paper proposes a hybrid neural network model based on CNNand LSTM.By preprocessing data through the sliding time windowalgorithm，themodelcombins theadvantagesofCNNinfeatureextractionwithLSTM'scapabilityintime-seriesdata procesing，achieving multi-categoryanomalydetectiononthe SWaTdataset.Experimentalresults demonstrate that theCNNLSTM modeloutperforms traditionalCNNmodelsandothercomparativemodels inindicatorssuchasaccuracyandrecall， while maintaining lowcomplexityandfastdetectionspeed，providing anefficient solutionforthesecurityprotectionof Industral Control Networks.

Keywords： Industrial Control System;intrusiondetection;Convolutional Neural Networks;Long Short-Term Memory; sliding time window

0 引言

工业控制系统（ICS）广泛应用于先进制造、交通运输以及城市设施等国家重要领域，其安全性对于保障国家的稳定和安全具有至关重要的意义。（剩余8346字）