Deep learning has emerged as a transformative approach for cybersecurity in Software-Defined Networking (SDN) environments. This paper presents a comprehensive deep learning framework for real-time Distributed Denial-ofService (DDoS) detection in SDN architectures serving consumer electronics ecosystems. We employ an optimized Deep Neural Network (DNN) that learns complex traffic patterns through multiple hidden layers, achieving superior performance: 99.19% accuracy, 99.15 % precision, 99.23 % recall, 99.19 % F1-score, and 0.9965 AUC-ROC. Our comprehensive evaluation compares the DNN against nine traditional machine learning classifiers using stratified 5-fold cross-validation with rigorous statistical significance testing. The DNN surpasses XGBoost (98.03 %) and SVM (97.20 %) with statistical significance ($p<0.05$). Our model demonstrates real-time feasibility with lightweight architecture ($\mathbf{8. 5 ~ M B}, \mathbf{1, 8 9 7}$ parameters), enabling $\mathbf{0. 3 4 ~ m s}$ inference per sample on SDN controllers. Cross-dataset validation across CICDDoS2019, CICIDS2017, and NSL-KDD reveals 410 % degradation. Adversarial robustness testing identifies $\mathbf{122 3 \%}$ vulnerability to FGSM attacks. We propose a two-tier detection architecture leveraging decision trees for fast screening and DNNs for deep analysis. This work validates deep learning as the next-generation intelligent security layer for SDN-based consumer electronics ecosystems.
Paper
Full text
AI-Driven Deep Learning for Real-Time DDoS Detection in Software-Defined Networks
Semantic Scholar · 2026
Abstract
Deep learning has emerged as a transformative approach for cybersecurity in Software-Defined Networking (SDN) environments. This paper presents a comprehensive deep learning framework for real-time Distributed Denial-ofService (DDoS) detection in SDN architectures serving consumer electronics ecosystems. We employ an optimized Deep Neural Network (DNN) that learns complex traffic patterns through multiple hidden layers, achieving superior performance: 99.19% accuracy, 99.15 % precision, 99.23 % recall, 99.19 % F1-score, and 0.9965 AUC-ROC. Our comprehensive evaluation compares the DNN against nine traditional machine learning classifiers using stratified 5-fold cross-validation with rigorous statistical significance testing. The DNN surpasses XGBoost (98.03 %) and SVM (97.20 %) with statistical significance ($p<0.05$). Our model demonstrates real-time feasibility with lightweight architecture ($\mathbf{8. 5 ~ M B}, \mathbf{1, 8 9 7}$ parameters), enabling $\mathbf{0. 3 4 ~ m s}$ inference per sample on SDN controllers. Cross-dataset validation across CICDDoS2019, CICIDS2017, and NSL-KDD reveals 410 % degradation. Adversarial robustness testing identifies $\mathbf{122 3 %}$ vulnerability to FGSM attacks. We propose a two-tier detection architecture leveraging decision trees for fast screening and DNNs for deep analysis. This work validates deep learning as the next-generation intelligent security layer for SDN-based consumer electronics ecosystems.