Optimization of Incremental Network Intrusion Real-Time Detection Algorithm Based on Computer Data Simulation

Abstract

High-quality and representative network intrusion data is the basis for building a network intrusion detection system. Given that traditional algorithms struggle to satisfy the demands of real-time detection, this paper constructs a network intrusion data simulation method based on LDMs, and combines it with an improved LST model to perform incremental real-time network intrusion detection. The experiments demonstrated that the LST model outperformed the comparison models in key indicators such as detection rate, precision, recall, and F1-score, and showed significant advantages in running time. After the LST model converged, its accuracy, recall, and F1-score were 0.983, 0.971, and 0.958, respectively. When dealing with advanced persistent threat attacks and zero-day attacks, the detection rate reached 97.4% and 96.8%, and the running time was 160.3 s and 155.2 s, respectively. The LST model not only achieved high accuracy in detecting complex network intrusion behaviors, but also has obvious advantages in computational efficiency. This paper provides an effective approach for improving network security, helps to promptly discover and prevent potential network threats, and protects the stable operation of network systems.