On The Research Of Edge-cloud Synergy Integrated Dynamic Security Decision-making Method For Industrial Cyber-physical Systems

Industrial Cyber-Physical Systems(ICPSs)are the transformation and upgrade of Industrial Control Systems(ICSs)with deep integration of information and communication technologies.They are Safety Critical Systems(SCSs),and it is essential to ensure their information security.With the introduction of new technologies such as cloud computing and big data,ICPSs are developing in the direction of intelligence,but their existing security issues have become more complicated.At present,a lot of security researches are focused on network systems and ICSs,which lack adaptability to industrial system upgrades.Therefore,it is necessary to construct a cyber-security protection system suitable for ICPSs.Security decision-making is the key part of the protection system,which has drawn increasingly more attention.Nevertheless,current security measures lack active decision capabilities to defend against Advanced Persistent Threat(APT).Based on the in-depth analysis of the security decision-making requirements in cloudintegrated ICPSs,an edge-cloud synergy framework for security decision-making is designed and a dynamic decision-making method is proposed under this framework.The edge-cloud synergy framework combines the resource advantages of cloud computing,such as massive storage and computing,with the rapid service response capability of the edge computing nodes,and is designed according to the scheme of “AI Training in the Cloud and AI Inference at the Edge”.Under this framework,the dynamic security decision-making problem of ICPSs is studied,and a risk-based security decision-making method is proposed,which mainly studies the risk assessment and security decision-making.Risk assessment is composed of two parts: Bayesian network-based probabilistic reasoning and asset evaluation based on fuzzy comprehensive evaluation(FCE)method.The product of these two parts is the quantitative result of risk assessment.The reasoning process and results of risk assessment are important foundations of the generation of attack-defense strategy and the payoff quantification of stochastic game model.In addition,the Bayesian network parameters are learned based on the proposed Map-Reduced EM algorithm,which is driven by big data to improve model accuracy.Security decision-making is to generate the optimal defense strategy for ICPSs,which is achieved by establishing a two-player non-zero and non-cooperative stochastic game model and designing a quantified payoff model.At the same time,based on proposed Q-learning algorithm in clouds,the stochastic game model is trained to solve the problem that the state transition parameters are difficult to determine accurately,and edge-cloud intelligent collaboration is realized to improve the dynamic adaptability of the model.Finally,the semi-physical simulation platform for ICPSs is built with hardware-in-theloop.Based on this platform,the security decision-making method is implemented and verified.A series of experiments,such as risk assessment,decision-making,and edge-cloud synergy,have verified the effectiveness and applicability of the proposed method.The research in this thesis shows that the scheme of edge-cloud synergy and model-data drive are of positive significance to the security decision-making and even the cyber-security protection of ICPSs.