Advancing Intrusion Detection in Software-Defined Networks
Atul Kumar Agrawal, Bharat Bhushan, Himani Sharma, Alaa Ali Hameed, Akhtar Jamıl · 2025
Although software-defined networks (SDNs) provide revolutionary configuration and flexibility, they are becoming more vulnerable to sophisticated security risks. This research addresses the difficulties of SDN systems by introducing a novel intrusion detection framework that combines Deep Reinforcement Learning with Graph Convolutional Network. While DRL guarantees adaptability to changing threats, GCNs improve the system’s ability to capture topological dynamics by utilizing the graph-structured data that is inherent in SDNs. Compared to more conventional models like Deep Neural Networks (DNN) and GRU-RNN, the suggested system performs better at detecting intrusions and uses the NSL-KDD dataset for training and evaluation. With precision, recall, and F1-score over $95 \%$ and an accuracy of $\mathbf{9 5. 8 5 \%}$, the framework demonstrates its efficacy in protecting SDNs from ever-changing cyberthreats. This research represents a breakthrough in intrusion detection techniques by offering a reliable and scalable way to improve SDN security.