摘要
工业控制系统(ICS)由原来的封闭隔离状态逐渐开放,导致工业控制设备出现故障或损坏。根据ICS业务逻辑稳定性的特点,提出一种新颖的ICS入侵检测模型,以确保系统数据和业务逻辑安全可靠。在边缘计算3.0的架构下构建边云协同的入侵检测模型;改进传统的卷积神经网络(CNN)算法,利用深度可分离卷积建立Mobile Net模型实现入侵检测业务的仿真。仿真结果表明,该模型的准确率提升到98%以上,训练时间是传统CNN模型的50%,更适合边缘端使用。
Industrial control system(ICS)is gradually opened up from the original closed and isolated state,which leads to failure or damage of industrial control equipment.According to the stability of ICS business logic,a novel ICS intrusion detection model is proposed to ensure the security and reliability of system data and business logic.The edge cloud collaborative intrusion detection model was constructed under the framework of edge computing 3.0;we improved traditional convolution algorithm of neural network(CNN)using depth separable convolution Mobile Net model to establish for intrusion detection business simulation.Through the simulation,the precision rate of our model is increased to more than 98%,training time is 50%of the traditional model of CNN,which is more suitable for the edge end.
作者
陈思
吴秋新
张铭坤
安晓楠
龚钢军
刘韧
秦宇
Chen Si;Wu Qiuxin;Zhang Mingkun;An Xiaonan;Gong Gangjun;Liu Ren;Qin Yu(Beijing Information Science and Technology University,Beijing 100192,China;Beijing Power System Information Security Engineering Technology Research Center in Energy Industry,North China Electric Power University,Beijing 102206,China;Beijing Excellent Network Security Technology Co.,Ltd.,Beijing 102206,China;Institute of Software,Chinese Academy of Sciences,Beijing 100190,China)
出处
《计算机应用与软件》
北大核心
2020年第11期280-285,333,共7页
Computer Applications and Software
基金
国家自然科学基金面上项目(61872343)
国家重点研发计划项目(2018YFB0904900,2018YFE0904903)。
关键词
边缘计算
云计算
入侵检测
深度可分离卷积
Edge computing
Cloud computing
Intrusion detection
Depth separable convolution
