Abstract & Overview
Abstract— Software-Defined Networking (SDN) offers centralized control and programmability but remains vulnerable to distributed denial-of-service (DDoS) attacks. In this paper, we present a novel stacked ensemble framework that integrates Feature Importance Analysis (FIA) and Synthetic Minority Over-sampling Technique (SMOTE) to enhance detection performance on the CICIDS2019 dataset. Our method comprises six main stages: data preprocessing, feature extraction, feature selection, data balancing, ensemble training, and SDN simulation. On extensive experiments, the proposed approach achieves 99.92% accuracy, 98.7% recall, and 0.08% false positive rate, outperforming existing state-of-the-art methods. The results demonstrate the model’s robustness and feasibility for deployment across diverse SDN controllers