DNN7: An Efficient Network Intrusion Detection Model
Hao Li, Xiangbo Wu, Yangyang Zhao, Xiaopeng Yang, Erlu He, Zhe Jia · 2022
With the explosive growth of network traffic, network intrusion detection has become an increasingly important part of network security field. In view of the frequent emergence of new attack types, it is necessary to design an efficient network intrusion detection system (NIDS) model. However, an realistic problem for current NIDS is the balance between the accuracy and real-time performance in massive traffic data scenarios. In this paper, we propose an improved deep neural network model named DNN7, which contains 7 hidden layers to better capture data features. Furthermore, batch normalization layer is introduced to reduce the change of the hidden layer data distribution and improve the stability of the network. Experiments on UNSW-NBIS indicates that DNN7 has excellent performance. Compared with the DNN3 model, the accuracy is increased by 20%. It can save a lot of time compared to other complex models with deep layers. DNN7 has a simple structure and less overhead, and is more suitable for real-time analysis and detection of massive traffic in real scenarios.