Research On Intrusion Detection Based On Convolutional Neural Network

As an important part of the network security field,intrusion detection uses various machine learning algorithms to detect unknown types of network attacks,thereby effectively ensuring network security.Convolution Neural Network(CNN)can process higherdimensional data and can automatically extract data features.Based on the above advantages,a convolutional neural network-based intrusion detection model is proposed.The research content mainly includes:First,the convolutional neural network is analyzed,and the intrusion detection data dimension is lower than the picture dimension.The classic neural network model Le Net-5is used to establish an intrusion detection model,and the process and specific calculations of the model to detect intrusion behavior information are introduced.Method.Aiming at the problem that the Le Net-5-based model in the intrusion detection data set KDDCUP99 has poor classification results due to the network connection and the sparsity of the data,it is proposed to improve the single-layer convolution in the Le Net-5 model to double-layer convolution.,And improve the model parameters according to the characteristics of the intrusion detection data set,thereby improving the model's ability to learn the characteristics of the intrusion detection data.The simulation results show that the accuracy of the improved model is increased by 3.64%,and the false alarm rate and false alarm rate are reduced by1.24% and 4.22%,respectively.Secondly,analyze the distribution characteristics of the training set and test set data in KDDCUP99,and perform preprocessing such as digitization,standardization,normalization,and de-redundancy on the data set.Aiming at the problem that the model's ability to classify minority classes is not strong,it is proposed to use the combined with oversampling and undersampling(COU)method in the training set to process the training set to enhance the classification ability of the model.In order to further improve the intrusion detection effect and improve the instability,the genetic algorithm(GA)and the convolutional neural network are combined,and the global optimization ability of the genetic algorithm is used to obtain the optimal initialization weight of the convolutional neural network.The intrusion detection model COU-ICNN-GA based on improved convolutional neural network is obtained.The simulation results show that the accuracy of the proposed COU-ICNN-GA model is 6.85%higher than that of the Le Net-5 based intrusion detection model,and the false positive rate and false negative rate are reduced by 2.43% and 7.92%,respectively.Finally,by comparing with other traditional algorithms and their improvements,the superiority of the proposed intrusion model based on COU-ICNN-GA is proved.Figure 31;Table 18;Reference 50...