Abstract The seminar offers an extensive examination of real-time threat detection in Software-Defined Networking (SDN) through Reinforcement Learning (RL), highlighting the shift from static deep learning models to adaptive, autonomous security frameworks. Traditional networks use distributed control, which makes them more complex an‎d less secure. On the other han‎d, SDNʹs centralized architecture makes them easier to construct but also makes them more vulnerable to attacks, especially against the controller. The research underscores the growing significance of Deep Reinforcement Learning (DRL), particularly architectures such as Deep Q-Networks (DQN), Double DQN (DDQN), an‎d Actor–Critic models, in mitigating dynamic, high-dimensional network threats, including Distributed Denial of Service (DDoS) attacks. It critically examines hybrid Deep Reinforcement Learning (DRL) frameworks that amalgamate Graph Convolutional Networks (GCN) with Explainable AI (XAI) to enhance topological awareness an‎d decision transparency. The conference also points out important research gaps, such as the lack of stan‎dardized SDN testbeds, the restricted ability to integrate telemetry across layers, an‎d the need for greater work on secure, understan‎dable online learning. These shortcomings make it hard to use RL-based systems in real-world networks. To overcome these constraints, the seminar suggests a forthcoming thesis entitled “A Lightweight Actor–Critic Framework for Real-Time DDoS Detection in SDN Controllers Utilizing Adaptive Reward Shaping.” This suggested method aims to improve detection accuracy an‎d latency, allowing SDN controllers to have defense mechanisms that work in real time, use few resources, an‎d are easy to understan‎d. This is part of a larger goal of creating autonomous, self-defending network infrastructures.

2/18/2026 12:00:00 AM

89793

1404/11/30

Real-time threat detection in SDN using reinforcement learning (RL)

Software-Defined Networking (SDN) , Reinforcement Learning (RL) , Deep Reinforcement Learning (DRL) , Distributed Denial of Service (DDoS), , Real-Time Threat Detection.