تحسين عملية الاسترداد من اختراق الشبكة في SDN باستخدام التعلم المعزز الاستباقي

Abstract

With the rapid evolution of Software-Defined Networks (SDN), the need for efficient network intrusion recovery strategies has become critical to ensuring service continuity and minimizing the impact of cyberattacks on network performance. This study proposes an innovative Reinforcement Learning-Based Network Intrusion Recovery (RLNIR) approach, which dynamically analyzes network traffic and instantly determines optimal alternative paths upon intrusion detection. This ensures minimal recovery time and improved bandwidth allocation for affected network flows. The proposed model was evaluated against several conventional and advanced strategies, including the ML approach in baseline, Fast Rerouting (FRT), proactive approach, and MLBNIR (Machine Learning-Based Network Intrusion Recovery). Experimental results demonstrate that RLNIR outperforms all existing approaches, reducing the network recovery time to 8 milliseconds, compared to 55 milliseconds in traditional methods. Additionally, it achieves an enhanced bandwidth allocation of 900 Mbps, the highest among all compared techniques. These findings highlight the potential of reinforcement learning as a future-proof solution for network intrusion recovery in SDN, enabling faster response to cyber threats, improved security flexibility, and dynamic resource allocation.