Attack detection and classification of smart grid based on residual network 50-convolutional block attention mechanism

To solve the problem that distributed denial-of-service attacks hinder the normal operation and affect stability of smart grid, a detection and classification method based on improved residual network and attention mechanism is proposed. Based on Residual Network 50 (ResNet50), the Convolutional Block Attention Mechanism (CBAM) is introduced to establish the detection model called ResNet50-CBAM. This model is suitable for the detection and classification of distributed denial of service attacks in interactive traffic with many features and a large amount of computation during training, and solves the gradient disappearance, gradient explosion and “degradation phenomena” existing in ResNet50. The experimental results show that in the multi-classification task such as detecting and classifying of 11 attack types, the overall accuracy is significantly improved.