基于MobileViT的轻量型入侵检测模型研究OA北大核心CSTPCD

In view of the data imbalance on the training of neural network model and the large number of model parameters in intrusion detection,an intrusion detection model based on improved MobileViT is proposed.The ANOVA(analysis of variance)is used to extract the features with great impact on the detection results,and the extracted features are converted into image-based data and input into the MobileViT network.For the attack traffic with a small proportion,the focus loss function is used to adjust the loss contribution of the attack traffic adaptively,so that the model can focus more on the unbalanced attack traffic.In order to solve the problem of neuronal death,the GeLU activation function is used to replace the ReLU6 activation function of MV2 in the MobileViT network to accelerate the convergence speed of the model.The results of experiments show that the improved MobileViT model has only 5.67 MB of parameters.It has the least parameters in comparison with Shufflenet and Mobilenet.The accuracy rate,recall rate and F1 score of the model reach 98.40%,96.49%and 95.17%,respectively.