Network security threat situation assessment by integrating deep learning and parallel feature extraction networks

To improve the accuracy and reliability of network security threat assessment, timely response and handling of complex network attacks are necessary. The study investigated feature extraction and threat assessment in the network. Then the concept of long short-term memory was introduced as a means of enhancing feature extraction from a single parallel network. Finally, a novel network security threat assessment model combining deep learning and parallel feature extraction networks was proposed. The parallel feature extraction detection accuracy of this model is the highest, at 94%. The false alarm rate on the NSL-KDD dataset is 9.23%, the false recognition rate is 7.21%, and the false alarm rate is 6.20%. Therefore, this study effectively improves the real-time responsiveness and accuracy of network security threat assessment. The research provides accurate, efficient, and practical solutions for network security attacks and defence.