Network Intrusion Detection Method Based on Hybrid Improved Residual Network blocks and Bidirectional Gated Recurrent Units

• Main Authors: Kang, C. (Author), Ting, Y. (Author), Xiao, Y. (Author), Yu, H. (Author)
• Format: Article
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2023
• Subjects: Bidirectional gated recurrent units; Feature extraction; Improved auto encoder; Logic gates; Network intrusion detection; Neural networks; Residual networks block; Residual neural networks; Time series analysis; Training data
• Online Access: View Fulltext in Publisher; View in Scopus

 With the rapid development of network technology, network intrusion detection plays a vital role in network security. In the era of big data, a large amount of network data is generated in the network all the time. Traditional detection methods do not achieve high accuracy and need to take a long time to detect network data. Therefore, how to improve the efficiency and accuracy of detection has become a hot topic of current research. Since each network traffic data has both spatial and temporal characteristics, this paper proposes a hybrid network classifier consisting of improved residual network blocks and bidirectional gated recurrent units. Before inputting the classification network, the feature dimensionality of the network data is first reduced using an improved autoencoder, and then the processed network data is detected using the constructed hybrid network classifier. In this paper, the proposed research approach is justified using official experimental datasets in the field of network detection (NSL-KDD and UNSW-NB15). The experimental results show that the proposed method in this paper achieves a higher accuracy of 93.40% and 93.26% on the datasets of NSL_KDD and UNSW_NB15, respectively, compared with the known detection methods. Author