A Comprehensive Hybrid Intelligence Approach Using Machine Learning and Deep Learning for DDoS Attack Detection in Software-Defined Networks
Year: 2026
Type: article
Abstract: Software-Defined Networks (SDNs) are increasingly vulnerable to sophisticated Distributed Denial of Service (DDoS) attacks, particularly in high-speed 5G/6G environments where traffic surges can overwhelm centralized controllers. To address this challenge, this paper proposes a hybrid, explainable deep learning framework for real-time DDoS detection and mitigation in SDN environments. The proposed system integrates multiple learning paradigms, including SVM-RF, Gradient Boosting–XGBoost, CNN-LSTM, and Transformer-based ensembles, to capture both spatial and temporal attack patterns. To improve model efficiency and robustness, Boruta feature selection is applied to eliminate redundant flow-level attributes derived from entropy-based statistics such as packet inter-arrival times and byte-per-flow ratios. In addition, SHAP-based explainability is incorporated to enhance transparency in detection decisions. The framework is evaluated using CIC-DDoS2019, CIC-DDoS2023, and SDN-NU datasets under SYN, UDP, and ICMP flood attack scenarios emulated in ONOS- and Ryu-based SDN testbeds. Extensive 10-fold cross-validation results demonstrate strong performance, achieving accuracies of 97.05% (SVM-RF), 98.82% (GB-XGBoost), 99.06% (CNN-LSTM), and 98.85% (Transformer-RF), with F1-scores exceeding 99% on imbalanced traffic distributions. Compared to single-model baselines, the proposed ensemble improves recall by 15–20% for zero-day attack detection while reducing false positives to 0.3%. System-level evaluation in Mininet-based 6G topologies shows a 52% reduction in detection latency and a 45% reduction in CPU utilization. Furthermore, TensorRT optimization enables sub-millisecond inference on edge devices such as NVIDIA Jetson, supporting real-time deployment in distributed SDN controller architectures. Overall, the proposed approach provides a scalable, efficient, and interpretable solution for strengthening SDN security against evolving DDoS threats in next-generation networks.
Author: Jayagopi G
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7218114110
