Imbalanced data handling in multiclass distributed denial of service attack detection using deep learning

Abstract

In data analysis, imbalanced datasets are a frequent issue, where classes in a dataset have an uneven distribution, which can lead to poor performance in machine learning (ML) and predictive modeling. In this study, we analyze distributed denial of service (DDoS) attacks at the application layer. Three primary strategies are studied in this study to address the issue of data imbalance in multiclass techniques: random oversampling (ROS), random undersampling (RUS), and the use of class weights. A model using a deep learning (DL) technique has been proposed in this paper to be trained and tested for DDoS attack detection. Based on the results obtained and presented in this paper, it is observed that RUS outperforms class-weight and ROS in multiclass settings in terms of resolving imbalanced data when implemented with the deep learning-based DDoS attack detection model.