جاسم هادي

Software-Defined Networking (SDN) changes the way networks are managed by putting all control in one place. However, this architecture has a big flaw: the controller is a single point of failure that may be overloaded by Distributed Denial of Service (DDoS) assaults. Conventional mitigation strategies frequently depend on static, reactive approaches that find it challenging to adjust to the evolving characteristics of contemporary multi-vector threats. This thesis presents an Adaptive Hybrid Metaheuristic Framework aimed at converting SDN security into an autonomic, self-optimizing system. The platform includes a hybrid Deep Learning detection module that uses Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, an‎d Attention mechanisms to accurately extract spatial-temporal aspects of malicious data. When an issue is found, a new metaheuristic optimization engine that uses Particle Swarm Optimization (PSO) an‎d Genetic Algorithms (GA) works in real time to find the best way to choose a mitigation plan that balances security efficacy with operational cost an‎d latency. A closed-loop feedback system is very important because it constantly changes the weights of strategies based on how well they worked after mitigation. The frameworkʹs strength was proven by experiments with the synthetic CIC-DDoS2019 dataset, which showed a detection accuracy of 86.58% an‎d a high precision of 90.55%. The system meets real-time requirements with an average total response time of 122.46 ms an‎d keeps 92.5% of legal traffic throughput during active attacks. These results prove that combining metaheuristic optimization with deep learning makes SDN far more resistant to advanced volumetric an‎d application-layer threats.

: شبكه‌هاي نرم‌افزارمحور (SDN)، , مقابله با DDoS , ، يادگيري عميق تركيبي , ، بهينه‌سازي فراابتكاري

Software-Defined Networking (SDN), , DDoS Mitigation , Hybrid Deep Learning , Metaheuristic Optimization

Jasim Hadi

Professor Javad Vahidi