Deep-learning-based intrusion detection methods

Abstract

With the increasing number and categories of cyber-attacks in the era of big data, intrusion detection techniques are constantly updated and optimized. In order to address the shortcomings of traditional intrusion detection methods and single neural network models in intrusion detection, this paper proposes a deep learning-based intrusion detection method. By combining a one-dimensional convolutional neural network with a bidirectional gated recurrent unit, a new CNN-BiGRU network model is formed to fully extract intrusion detection data features and improve the accuracy of intrusion detection. Using the public dataset CIC-IDS2017, multiple sets of comparison experiments were conducted on the proposed method in this paper in the same environment. Firstly, the ablation experiments yielded that in terms of detection performance, the accuracy of this paper's method improved from 99.50% and 99.34% to 99.58% compared with CNN and BiGRU-based models respectively. In terms of running time, the detection time of this model is reduced from 1921s to 907s compared with that of BiGRU, demonstrating the good detection performance of this method. Finally, through comparison tests, it is verified that the intrusion detection method proposed in this paper has better detection performance compared with other methods.