Beihang Team Unveils BGA Framework: Detecting Encrypted Attacks in 0.28ms with a 'Neural Filter'

A research team led by Professor Hong Sheng at Beihang University published a study in IPM proposing a noise-immune neural distillation framework called BGA. The framework tackles attention dilution in encrypted traffic by using WGAN-GP data augmentation, ANOVA feature selection, and adaptive gated multi-head attention. In industrial IoT scenarios, BGA achieves high detection accuracy and low latency, with inference as fast as 0.28ms, offering a robust solution for real-time security defenses.