Research On Anomaly Detection Method Of Physical Process Of Industrial Control System Based On FPGA

With the deep integration of informatization and industrialization,the industrial control systems is facing serious threats of network attacks.At the same time,due to the close coupling between the information layer and the physical layer,the network attack against industrial control system will penetrate into the physical layer and cause significant damages to the physical process.Therefore,protecting the physical process security of industrial control system is the key part of information security protection of industrial control system.Anomaly detection,as the core of physical process security protection,has been paid more and more attention by researchers.This thesis analyzes the anomaly detection requirements of the physical process of industrial control system.Aiming at the problems of low detection accuracy and poor real-time performance of the existing anomaly detection technologies,a general framework for anomaly detection of physical processes of industrial control system is proposed.The framework includes two parts: the design of the anomaly detection algorithm and the embedded implementation of the algorithm.The former designs the time series anomaly detection algorithm based on three-layer long short-term memory(LSTM)network,and the latter uses the parallelism and pipeline characteristics of field programmable gate array(FPGA)to accelerate the execution speed of LSTM in anomaly detection algorithm to meet the real-time requirements of anomaly detection.An anomaly detection algorithm for physical process of industrial control system is proposed based on deep learning technology.The algorithm consists of three steps: data processing,state prediction and anomaly identification.Firstly,the data quality is improved by data filtering and normalization.Then,the state prediction is realized based on LSTM network.Finally,according to the difference between the predicted value and the real value,the anomaly is identified by using one class support vector machine.The embedded implementation of anomaly detection algorithm is studied based on PYNQ(ARM + FPGA)development platform,and the design is carried out by combining software and hardware.In the hardware part,the LSTM accelerator is designed and implemented based on FPGA to accelerate the state prediction of anomaly detection algorithm.In the software part,the embedded implementation of the algorithm is completed through the data processing,hardware accelerator call and anomaly identification based on ARM processor.Finally,taking the multi-level water tank control system as the research object,the anomaly detection algorithm verification and embedded anomaly detection experiment are carried out.Experimental results show that the average anomaly detection rate of the proposed embedded anomaly detection method is 86.7%,the system power consumption is 2.03 W,and the average anomaly detection time is 6.7ms,which is 11.9 times faster than using only ARM processor,which verifies the effectiveness and real-time performance of the proposed method.