工业控制网络入侵检测的BP神经网络优化方法

BP neural network optimization method for industrial control network intrusion detection

针对工业控制系统入侵检测模型对各类攻击的检测率和检测效率不高的问题,提出一种Ada Boost算法优化BP神经网络的入侵检测模型.首先利用主成分分析法对原始数据集进行预处理,消除其相关性;其次利用Ada Boost算法对训练样本的权重进行不断调整,从而获得BP神经网络最优权重和阈值;最后再通过Ada Boost算法将BP弱分类器组合成BP强分类器,从而实现工业控制系统的异常检测.实验结果表明该方法在对各攻击类型的检测率和测试时间明显优于其他算法模型.

Aiming at the problem of low detection rate of typical attack types and low detection efficiency in industrial control system intrusion detection model,an intrusion detection model optimized by BP neural network based on AdaBoost algorithm is proposed.Firstly,the original data set is preprocessed by PCA to eliminate its correlation;Secondly,the AdaBoost algorithm is used to continuously adjust the weight of the training samples,thereby obtaining the optimal weight and threshold of BP neural network.Finally,the BP weak classifier is combined into a BP strong classifier by AdaBoost algorithm,and the anomaly detection of the industrial control system is realized.The experimental results show that the methods are superior to other algorithm models in detecting the attack types and test time.