摘要
针对工业控制系统容易遭受网络入侵威胁,进而影响工业控制系统安全性的问题,提出了一种结合生成对抗网络和深度神经网络的工业控制系统入侵威胁检测算法模型。该模型首先提出了一种样本均衡生成对抗网络,将反向传播神经网络(back propagation neural network,BPNN)作为分类器对入侵威胁进行分类,并通过蜻蜓优化算法实现对BPNN的改进。然后,结合开集识别和深度神经网络来实现对未知攻击的检测。最后,采用KDD数据集对模型的性能进行测试。实验结果表明,已知攻击的入侵威胁检测模型的准确率能够达到98%,F1值为0.947,召回率为0.975;未知攻击检测模型的精度为0.987,F1值为0.973,证明所提出的工业控制系统入侵威胁检测算法模型具有较高的检测精度,有效保障了工业系统的安全性。
To address the vulnerability of industrial control systems to cyber intrusion threats and the resultant security risks,this study proposes an intrusion threat detection model that integrates a generative adversarial network with a deep neural network.The proposed framework comprises three key components.Firstly,a sample-balanced generative adversarial network is developed to mitigate data imbalance issues.This SB-GAN employs a back propagation neural network(BPNN)classifier optimized through the dragonfly optimization algorithm for enhanced intrusion classification.Secondly,an open-set recognition mechanism combined with DNN architectures enables effective detection of unknown attack patterns.Finally,comprehensive evaluations on the benchmark KDD dataset demonstrate the model’s superior performance.Experimental results indicate that the proposed system achieves a detection accuracy of 98%with F1-score of 0.947 and recall of 0.975 for known attacks,while maintaining 0.987 precision and 0.973 F1-score for unknown threats.These results demonstrate the model’s high detection fidelity and its potential to significantly enhance industrial system security.
作者
胡智锋
孙峙华
HU Zhifeng;SUN Zhihua(Modern Education Technology Center,Wuhan Business University,Wuhan 430056,China;School of Smart Manufacturing,Jianghan University,Wuhan 430056,China)
出处
《控制工程》
北大核心
2025年第3期400-408,共9页
Control Engineering of China
基金
湖北省自然科学基金资助项目(2023AFB588)。
关键词
工业控制系统
生成对抗网络
网络入侵检测
深度神经网络
蜻蜓优化算法
Industrial control system
generating adversarial networks
network intrusion detection
deep neural networks
dragonfly optimization algorithm
作者简介
通信作者:胡智锋(1975-),男,湖北武汉人,硕士,高级实验师,主要从事计算机网络技术、软件工程、网络安全等方面的教学与科研工作(Email:hzf@wbu.edu.cn)。
