A Cloud-Based Intrusion Detection Data Security BSO-ResNet for Malicious Packet Detection for Malicious Network Congestion in 6G Networks

The combination of Black Swan Optimization (BSO) and Residual Networks (ResNet) forms a robust framework for detecting malicious packets within network traffic. ResNet’s deep learning architecture learns intricate patterns through residual connections that mitigate the vanishing gradient problem, enabling effective training of deep models. BSO enhances this framework by optimizing feature selection, reducing redundancy, and improving detection accuracy through lower false positive (FP) and false negative (FN) rates. The resulting system supports two complementary deployment modes: An accuracy-oriented configuration, which leverages a richer optimized feature set and deeper residual learning for security-critical environments, and a latency-oriented configuration, which uses a lightweight feature subset to achieve faster execution and reduced computational overhead for real-time congestion-attack detection in resource-constrained 6G scenarios. This dual capability makes the BSO–ResNet integration suitable for both high-assurance cloud security and ultra-low-latency edge environments.