Research of Network Intrusion Detection Based on Improved Seagull Optimization Algorithm with Deep Learning

In this paper, we study a network intrusion detection method based on deep learning combined with improved seagull optimization algorithm, which extracts the information traces inevitably generated during network intrusion by deep neural network and optimizes the parameters of deep neural network model by improved seagull optimization algorithm, so as to build an efficient network intrusion detection model. The traditional seagull optimization algorithm is improved and applied to the deep learning model hyperparameter optimization. For the shortcomings of the traditional seagull optimization algorithm with strong randomness of population initialization and easy to produce extreme individuals, a reverse learning method is presented to the initialization of the group. And a nonlinear convergence factor is used to enhance the convergence speed, thus improving the performance of the seagull optimization algorithm. The improved algorithm was demonstrated by using standard test functions, and the improved algorithm was used for parameter optimization of the deep learning model. To address the shortcomings of classical rule-based, host behavior analysis, and machine learning network traffic classification methods in performing network intrusion detection with less attention to the temporal correlation characteristics of samples, we propose to apply deep learning techniques to network intrusion detection, and design a network intrusion detection method based on gated cyclic units and multilayer perceptron, and also apply the improved seagull optimization algorithm to the method optimization of hyperparameters in the model, thus improving the performance of the model and achieving better network intrusion detection results on the NSK-KDD dataset.