Research on Intrusion Detection Method Based on PGoogLeNet-IDS Model

The current network environment generally presents a situation of high latitude and a massive amount of information. Intrusion Detection Systems not only need to improve detection rate but also enhance detection speed. Therefore, this paper applies the PGoogLeNet-IDS model to the Intrusion Detection System, which can effectively balance the relationship between detection rate and detection speed. First, the PGooglLeNet-IDS need to replace the Inception structure of GoogLeNet with the SE_DSC structure. Second, this model uses a cross-layer connection to the network layer of the model, which can reduce the loss of features in the training process. Finally, this model uses the Focus Loss Function to the model to solve the uneven distribution of sample data. This paper uses the NSL-KDD benchmark data set to verify the performance of the model. The experimental results show that PGoogLeNet IDS has been significantly improved in various performance evaluation indicators, and the training time of the model is also significantly shortened. The model proposed in this paper can be effectively detected in the current complex network environment.