Situation Awareness Technology of LeNet-5 Attack Detection Model Based on Optimized Feature Set

摘要

With the rapid development of China's industrial control system, in order to deal with its increasingly prominent security problems, this paper proposed an improved LeNet-5 convolution neural network in attack detection by establishing and optimizing the effective feature set to optimize the selection of feature data. And then, after calculating and extracting the feature value in convolution layer and pool layer, inputs the results into softmax classifier to achieve the detection of network attacks. At last, KDD CUP99 is used to test the proposed model. The experiment results show that the performance of the improved LeNet-5 attack detection model has a certain feasibility and works better than the traditional machine learning method, which can reduce the redundancy of data samples and improve the accuracy of attack detection.