重庆理工大学学报2025，Vol.39Issue(5)：85-92,8.DOI:10.3969/j.issn.1674-8425(z).2025.03.011
Transformer融合CNN-SRU的工业控制网络入侵检测方法
An industrial control network intrusion detection method integrating Transformer with CNN-SRU
摘要
Abstract
Current methods for industrial control network intrusion detection focus only on either local or global features of network traffic and the imbalance in network traffic data distribution leads to low accuracy of intrusion detection models.To address these issues,this paper proposes a novel method that integrates Transformer with Convolutional Neural Network-Simple Recurrent Unit(CNN-SRU)for industrial control network intrusion detection.Adaptive Synthetic Sampling Method(ADASYN)and Gaussian Mixture Model(GMM)are employed to oversample the minority class samples to achieve sample balance.The CNN-SRU captures the spatiotemporal local features of network traffic data,while the Transformer encoder part captures global connections for deep feature extraction.Our experiments on the NSL_KDD dataset show the overall accuracy of our model reaches 99.61%,higher than that of the compared neural network models.Experimental validation on the Mississippi State University gas pipeline control system dataset and the laboratory's oil and gas gathering and transportation full-process industrial attack-defense range indicates the overall accuracy reaches 98.58%and 96.89%respectively,demonstrating the validity and feasibility of our method in industrial control network intrusion detection.关键词
工业控制网络/入侵检测/Transformer/卷积神经网络/简单循环单元Key words
industrial control network/intrusion detection/Transformer/convolutional neural network/simple recurrent unit分类
计算机与自动化引用本文复制引用
史长鑫,宗学军,何戡,连莲,孙逸菲..Transformer融合CNN-SRU的工业控制网络入侵检测方法[J].重庆理工大学学报,2025,39(5):85-92,8.基金项目
辽宁省自然科学基金项目(2023-MSLH-273) （2023-MSLH-273）
辽宁省科学技术计划项目(2023JH1/10400082) （2023JH1/10400082）
辽宁省人工智能创新发展计划项目(2023JH26/1030008) （2023JH26/1030008）
辽宁省科技创新平台建设计划项目([2022]36号) （[2022]36号）
