Research on Network Intrusion Detection Techniques Based on Feature Selection Model and Recurrent Neural Network
Xiaomei Liu, Yujing Cao · 2023
The number of network attacks has also increased rapidly. Therefore, it is necessary to conduct in-depth research on network intrusion detection technologies. Compared with traditional intrusion detection systems, AI-based intrusion detection systems can better detect network traffic, with lower false positive and false negative rates. This paper first expanded the CIC-IDS-2017 dataset by adding two new attack forms. Secondly, the SMOTE algorithm was used to expand the minority samples in the dataset. Through the sequence forward selection algorithm based on decision trees, the features in the dataset were selected, improving the algorithm's efficiency without significantly affecting its accuracy. Finally, a multi-task network intrusion detection model was constructed by integrating the prediction results of the recurrent neural network model and the one-class support vector machine model to determine the final network traffic type. The system has been quite successful in achieving all of its initial goals at this stage of research. The system achieved a detection accuracy of over 90% for network traffic can predict unknown types of attack traffic. The real-time detection function allows the system to be applied to practical network traffic detection on a daily basis.