Research And Application Of Visual Optimization Technology In Intrusion Detection Of Industrial Control System

With the deep integration of the two technologies,the information security problem of industrial control systems is becoming more and more serious.Intrusion detection is the key link of information security protection of industrial control systems.However,the existing intrusion detection technologies have problems such as low accuracy and poor real-time performance.There is an urgent need for a technology that can effectively mine key information and potential patterns from massive data.Visual optimization technology can display massive data in a visual form,helping users to discover knowledge,analyze rules and optimize models based on expert experience,and improves the existing problems of intrusion detection.Firstly,according to the typical structure of industrial control system,this thesis analyzes the potential security threats,and then introduces the visualization optimization technology from three aspects of visual display,interaction and analysis technology,which leads to the requirement of visual optimization in intrusion detection of industrial control system,and proposes a visual optimization framework.Mainly using the "human in the loop" optimization idea,breaking the "black box" mode of traditional industrial control system intrusion detection.In the preprocessing stage(data domain),the data cleaning optimization data is mainly used through the scatter plot matrix;the key feature selection phase(feature domain),according to the characteristics of the data set,the information layer uses the mutual information and XGBoost algorithms to perform feature ranking.Combining information entropy theory and expert knowledge comprehensive analysis to select key features,the physical layer uses recursive feature elimination algorithm to automatically select key features;algorithm selection phase(algorithm domain),provides two different algorithms SVM and KNN to select,and supports algorithms the adjustment and optimization of the parameters of the super-superior;the evaluation stage(assessment domain),the comprehensive analysis of the classification performance is carried out by means of the classification comprehensive report and the confusion matrix,and the model is iteratively optimized according to the evaluation results.Based on the above scheme,this thesis designs and implements the industrial control system intrusion detection visualization optimization system,and uses different data sets in the information layer and the physical layer to carry out experimental verification and comparative analysis.The accuracy of the information layer test set is as high as 99.54%.It is superior to other thesis that use the same data set;the accuracy of the physical layer test set is as high as 99.24%,and the online test shows that the system can accurately detect the attack and the specific attack type.In summary,the effectiveness and practicability of the proposed scheme and design system are proved.