Traditional intrusion detection systems fail against zero-day attacks an‎d rely on outdated datasets like KDDCUP99, while AI tools like ChatGPT exhibit only 69% accuracy in network configuration — a critical risk for SDN security. This work introduces the first integrated framework that unifies an unsupervised NDAE for anomaly detection with a cooperative MADDPG agent for dynamic traffic steering, trained on the realistic UNSW-NB15 dataset an‎d deployed on a Mininet-Ryu SDN testbed. Unlike prior work, our system does not generate OpenFlow rules — it decides when an‎d where to reroute traffic based on learned behavior. eva‎luated across five key metrics (F1-Score, Latency, FPR, CPU Load, Detection Time), our AI-SDN Guardian achieves 95.7% F1-Score, 87ms latency, an‎d 1.8% FPR — outperforming stan‎dalone models an‎d setting a new benchmark for autonomous, adaptive SDN security.

10/29/2025 12:00:00 AM

88000

1404/08/08

Application of Artificial Intelligencein Software Defined Networking For Enhancing Network Traffic Management an‎d Intrusion Detection

AI , SDN , NDAE , MADDPG , Intrusion Detection , Traffic Management , Network Security