Network security situation prediction method based on KG and Parsimonious memory unit

With the continuous development of the Internet, the application scope of the Internet is constantly expanding, and network security issues are becoming increasingly prominent. To address long prediction time and low prediction efficiency in the current network security situation prediction model, this study first designs a KG of network security situation. Afterwards, the GCN and MLP are utilized to extract features from the data in the KG as data for the subsequent prediction model. It then uses the CBAM and PMU to predict the network security situation based on feature data. The test results demonstrated that during feature extraction, the feature extraction accuracy could reach 98.4%, and the prediction accuracy of the optimized PMU could reach 98.5%. Moreover, according to the designed model, its prediction accuracy could reach 96.7%, the prediction time only took 11.2 ms, and the MAE was only 0.081, which means the model's prediction effect is good. When the model predicted under ± 30% network traffic disturbance, large wide area network conditions, 20% equipment failure, and network virus attacks, its prediction accuracy was higher than 90%. The research method can predict the network security situation, thereby providing early warning information for managers and reducing network risks.