Network Intrusion Detection Based on VDCNN and GRU Fusion
Heyu Wang, Yinggang Li · 2023
The surge in network traffic has led to an increase in network attacks, and network security has received more and more attention. Intrusion detection is a very important part of network security. However, the accuracy of traditional intrusion detection models is not high. In order to improve the recognition rate of attack traffic. An intrusion detection model combining Very Deep Convolutional Neural Networks (VDCNN) model and Gated Recurrent Unit (GRU) model is proposed. Firstly, preprocessed the data by One-Hot encoding and Normalization. Secondly, data features are obtained through VDCNN and GRU: VDCNN network structure can extract high-dimensional data features and edge features of data, and GRU can obtain temporal correlation features between data; finally, combined two types of features to form new data features, which are transmitted to the Softmax classifier for classification and recognition as new input. The experiment uses two public datasets: KDD Cup 99. The experimental results show that the accuracy of intrusion detection on the dataset is 98.73 %, which proves that the model has good ability. Compared with other models, VDCNN-GRU improves at least 4.36%, to verify the recognition ability of the model.