DAD: Enhancing Multi-Class DDoS Attack Classification using Data Augmentation with DRCGAN

Currently, the use of deep learning based methods to detect DDoS attacks has a broad prospect, and these techniques can greatly improve the efficiency of intrusion detection systems. However, data-driven based techniques require a large quantity of training data support, and existing network traffic datasets often suffer from insufficient samples and unbalanced data distribution. In this paper, we introduce a DDoS attack detection method based on data augmentation. The method utilizes a Deep Residual Conditional Generative Adversarial Network (DRCGAN) to generate diverse and high-quality samples. The method mitigates the data imbalance problem and improves the performance of the DDoS attack detection model. Extensive experiments on the CICDDoS2019 dataset and the dataset based on DDoS attack family classification show that the proposed method improves the multi-type attack detection precision to 88.33% and 99.55%, respectively, which highlights its excellent detection performance.