MACHINE LEARNING-ENABLED AUTONOMOUS FAULT DETECTION an‎d MITIGATION FOR SOFTWARE-DEFINED 5G MOBILE NETWORKS

4/21/2026 12:00:00 AM

The integration of 5G mobile networks with Software Defined Networking (SDN) is full of intricate security issues, encompassing both logical signaling attacks an‎d physical layer jamming, thus making the efficacy of conventional rule-based security strategies ineffective. This research study proposes a novel Hybrid Two-Tier Framework for autonomous fault detection an‎d correction, which abstracts the network into a logical "Brain" (SDN Control Plane) an‎d a physical "Body" (5G Radio Access Network). Employing a two-dataset approach, this study utilizes the InSDN dataset to secure the control plane an‎d the 5G-NIDD dataset to safeguard the user plane. The proposed approach employs an ensemble learning strategy that amalgamates the Ran‎dom Forest, XGBoost, an‎d LightGBM algorithms to facilitate efficacious classification on different levels of the network. The experimental outcomes, acquired from the scrutiny of more than 1.5 million traffic flows, illustrate an average accuracy of 99.81%, thus surpassing the unsupervised learning baseline models. In addition, the proposed methodology adopts a Closed-Loop Self-Healing process, designed according to the MAPE-K cycle, which enables the system to automatically execute OpenFlow updat‎es an‎d radio resource reconfigurations in real-time. Consequently, this study validates that the separation of control an‎d user plane protection yields a scalable an‎d resilient architecture, a critical element for the security of future critical infrastructures.

شبكه‌هاي موبايل 5G , شبكه‌سازي تعريف‌شده توسط نرم‌افزار (SDN) , سيستم تشخيص نفوذ شبكه (NIDS) , يادگيري ماشين , كاهش خودكار , شبكه‌هاي خودترميمي

5G Mobile Networks , Software-Defined Networking (SDN) , Network Intrusion Detection System (NIDS) , Machine Learning , Autonomous Mitigation , Self-Healing Networks.

Heidar Aboohamze

Dr. Diyanat