Research On The Key Techniques Of Mimic-structure-based Heterogeneous Processing System

Industrial control system usually contains many heterogeneous processing nodes.As a service-oriented information processing system,it undertakes tasks of monitoring,data acquisition,control and calculation in specific scenarios.In recent years,the utilization of information technology in the industrial control system not only improves operational efficiency,but also increases the attack surface of the system itself,exposing the vulnerability and backdoors of many devices to the Internet.This change makes the industrial control system face many security challenges because it had a closed character and took reliability along with availability as the main design objectives.Self-manufacturing and vulnerability testing can not completely eliminate devices' vulnerability and backdoors.Existing defense technology,relying on prior knowledge or simple dynamic shifts,can not guarantee the credibility of the system based on these unsecure devices.Mimic defense techniques provide a solution to security problems in industrial control systems with DHR(Dynamic Heterogeneous Redundancy)structure.By voting on the output vectors of heterogeneous executors and triggering negative feedbacks to change defense environment,mimic defense techniques can achieve overall security of the system even under the condition that the executor "carries the virus".Therefore,discussions about mimic defense techniques in the field of industrial control have great practical significance.Based on the background of mimic security processor for industrial control system,this paper first gives a feasible method for evaluating the heterogeneity between executors in the processor.Then based on the results of heterogeneity evaluation,a real-time task scheduling problem is discussed under DHR architecture,and a voting scheme is given to solve the problem of common errors in multiple executors.The main work of this paper is as follows:1.To evaluate the heterogeneity between executors in mimic security processors,an evaluation method integrating software and hardware components is proposed.This method divides the software and hardware components of the executor by tree hierarchical structure,and selects qualitative or quantitative evaluation indicator of heterogeneity according to the features of each component.The weight values of the nodes in the tree are assigned by analytic hierarchy process.Finally,the heterogeneity between two executors are obtained by fusing multiple values of evaluation indicators with D-S inference and weighted calculation.An example is given to show the rationality and feasibility of the evaluation method.2.Based on the positive correlation between heterogeneity and the probability that the system does not have common errors,a redundant adaptive heterogeneity-first scheduling algorithm is proposed under DHR architecture.Considering the real-time constraints of tasks,the algorithm uses a global queue related to node selection to increase the fairness of scheduling and maximizes the average heterogeneity between executors with a greedy strategy.The algorithm also improves the schedulability with a conditional redundant task discarding strategy when the system is busy.Simulation results show that the mean value of the average heterogeneity of the proposed algorithm is 22.7% higher than that of the comparative algorithm under the condition of ensuring the guarantee ratio.3.To increase correctness of voting when executors output common errors,an improved fuzzy voting scheme based on the heterogeneity between the executors is proposed.Firstly,improved fuzzy rules are proposed according to the number of redundant executors along with related common error conditions.Secondly,heterogeneity information between executors is used to assist a second voting.Simulations prove the effectiveness of the proposed scheme,as well as its better results of reliability and availability indicators compared with other schemes.