融合LSTM-DNN的工业安全态势预测模型  

Industrial Security Situation Prediction Model Based on LSTM-DNN

作　　者：于雅洁 刘贤达 蒋启梅[5] 张博文 YU Ya-jie;LIU Xian-da;JIANG Qi-mei;ZHANG Bo-wen(Key Laboratory of Networked Control Systems,Chinese Academy of Sciences,Shenyang 110016,China;Shenyang Institute of Automation,Chinese Academy of Sciences,Shenyang 110016,China;Institutes for Robotics and Intelligent Manufacturing,Chinese Academy of Sciences,Shenyang 110169,China;University of Chinese Academy of Sciences,Beijing 100049,China;ChangHe Aircraft Industry(group)Corporation LTD.,Jingdezhen 333001,China)

机构地区：[1]中国科学院网络化控制系统重点实验室,沈阳110016 [2]中国科学院沈阳自动化研究所,沈阳110016 [3]中国科学院机器人与智能制造创新研究院,沈阳110169 [4]中国科学院大学,北京100049 [5]昌河飞机工业(集团)有限责任公司,江西景德镇333001

出　　处：《小型微型计算机系统》2023年第3期596-601,共6页Journal of Chinese Computer Systems

基　　金：国防基础科研项目(JCKY2020205B022)资助。

摘　　要：态势预测对于感知工控系统中的安全风险有着重要的作用.传统的态势预测模型往往会忽略工控系统中态势要素的时序性,难以准确对系统的安全态势进行预测.因此本文提出一种基于LSTM-DNN的工业网络安全态势预测模型,以提高传统态势预测模型的精确度.首先从海量数据中选取出与系统态势强相关的态势要素;接下来利用LSTM对提取的态势要素进行预测,得到未来的态势要素链;最后将提取出的态势要素链送入DNN模型中,预测系统未来的安全态势.实验表明,相较于传统的网络安全态势预测模型,该模型框架能够有效地预测未来的态势值;相比于其它算法,所提出的算法具有较高的预测精度.Situation prediction plays an important role in sensing the security risk of industrial control system.The traditional situation prediction model often ignores the time sequence of situation elements in industrial control system, so it is difficult to accurately predict the security situation of the system.Therefore, this paper proposes an industrial network security situation prediction model based on LSTM-DNN to improve the accuracy of traditional situation prediction model.Firstly, the situation elements strongly related to the system situation are selected from the mass data;Next, LSTM is used to predict the extracted situation elements, and the future situation element chain is obtained;Finally, the extracted situation element chain is sent into the DNN model to predict the future security situation of the system.Experimental results show that compared with the traditional network security situation prediction model, the model framework can effectively predict the future situation value;Compared with other algorithms, the proposed algorithm has higher prediction accuracy.

关 键 词：工控网络 态势评估 态势要素链 长短时记忆网络 全连接神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]