CBAM-EANet: Convolutional Block Attention Module Enhanced Efficient Network for Detection on Distributed Denial of Service Attacks

With the emergence of numerous DDoS attack methods, traditional detection techniques can no longer accurately and swiftly detect DDoS attacks. Furthermore, deep learning-based traffic image classification models often become overly complex in their pursuit of higher accuracy. In the paper, we propose a deep learning model-based method for the visual detection of DDoS malicious traffic. In response to the issue of model complexity, this article has made improvements based on the EfficientNet model, especially in resource constrained environments. This article introduces CBAM into the original model to improve its ability to process global information. By replacing the loss function in the final Softmax layer of the EfficientNet model, the ability to extract image features for binary classification problems has been enhanced, resulting in improved model performance. The experimental results show that this method can effectively identify various DDoS attacks. Compared with the EfficientNet model, this method has improved accuracy, training time, and single image prediction time.