Network Security Situation Prediction Model Based on Neural Network and Gated Recurrent Unit

ABSTRACT With the development of Internet technology, network security (NS) has become an important issue in current information technology. To enhance the prediction precision of NS situations, this study designs a prediction model based on Convolutional Neural Network (CNN) and residual Gated Recurrent Units (GRUs). This study first combines principal component analysis and probabilistic neural networks (PNNs) to construct a NS situation assessment model, improving the computational efficiency and accuracy. On this basis, data features are extracted through convolutional neural networks, and an improved gating loop unit is used to solve the issue of information loss in transmitting situational information. The results demonstrated that the loss values of the proposed model on the CICIDS 2018 dataset and the UNSW‐NB15 dataset were 0.15 and 0.17, respectively. Compared with the other two models, the proposed model had significant advantages in convergence speed and prediction accuracy. For example, the area under the curve (AUC) value of this model was 0.9512 in DDoS attack scenarios and 0.9678 in normal traffic scenarios, demonstrating strong real‐time processing capabilities. This model also outperformed other models in terms of memory usage and computing speed. In attack testing, the memory usage remained stable between 2100 and 2300 MB, demonstrating higher efficiency and lower memory requirements, providing a new solution for situation prediction in NS.