Anomaly Detection Method Of Industrial Control System Integrating Quantum Particle Swarm Optimization And Machine Learning

The industrial control system provides a stable and convenient execution environment for industrial production.With the advancement of policies in various countries and the development of a new generation of information technology,the industrial production environment is gradually interconnected with the outside world.This interconnection breaks the previous closed situation of the industrial control system,causing the industrial control system to face more threats from the outside.The control system itself lacks the design flaw of security protection,and the security problems faced by the industrial control system are becoming more and more serious.At present,the existence of abnormal data in industrial control systems affects the stability and security of the system,and the unbalanced characteristics of abnormal data in the system also bring difficulties to the detection of abnormal data.One-class support vector machine is an unsupervised learning algorithm in the field of machine learning.Since only one type of data can complete the training of its model,it is widely used in practical imbalanced classification problems.Therefore,this paper introduces machine learning.A class of support vector machines in Industrial Control Systems detects abnormal data.Aiming at the difficulty in setting the optimal parameters of a class of support vector machines,this paper combines quantum particle swarm optimization with quantum particle swarm optimization,and proposes an anomaly detection method that integrates quantum particle swarm optimization and a class of support vector machines.In addition,the initialization and local optimal problems of quantum particle swarm also affect the detection accuracy.To solve these problems,this paper further proposes an anomaly detection method that combines the improved quantum particle swarm algorithm and a class of support vector machines.The work and innovation of this paper are as follows:(1)An anomaly detection method that integrates quantum particle swarm optimization and a class of support vector machines is proposed.First,in view of the inconsistency of dimensions among various feature attributes of industrial control data and the generally high dimension of data features,the normalization method and the principal component analysis method were adopted to normalize and reduce the dimension of the data.The key parameter setting problem of support vector machinelike,using quantum particle swarm algorithm to select parameters intelligently.Experiments are carried out on the proposed method on the industrial control data set published by the University of Mississippi,and the results show that the fusion of quantum particle swarm is better than a class of support vector machines that fuses particle swarm optimization,differential evolution,genetic algorithm and Bayesian algorithm.A class of support vector machines of the algorithm is more effective in detecting abnormal data of industrial control,and the detection accuracy of abnormal data of industrial control on the public data set reaches 94.2%.(2)An anomaly detection method that integrates improved quantum particle swarm optimization and a class of support vector machines is proposed.Firstly,for the problem of uneven distribution of the initial particle population in the search space and low quality of the quantum particle swarm optimization,this paper uses the ergodic and random characteristics of chaos in the quantum particle swarm optimization to improve the quality of the initial population;then,for the quantum particle The swarm algorithm is easy to fall into the local optimal problem in the process of intelligently selecting the optimal parameters.In this paper,the dynamic nonlinear contraction and expansion coefficient is used to improve the search performance of the algorithm.The test results on the public data set show that a class of support vector machines integrated with the improved quantum particle swarm algorithm has better anomaly detection performance,and the detection accuracy of the anomaly detection model constructed based on this method reaches 96.2%.