Adversarial deep reinforcement learning for autonomous cyber defense in software defined networks
| dc.contributor.author | Borchjes, Luke David | |
| dc.date.accessioned | 2025-09-25T10:15:20Z | |
| dc.date.available | 2025-09-25T10:15:20Z | |
| dc.date.issued | 2024 | |
| dc.description.abstract | The rapid advancement of technologies such as IoT devices, Internet-based AI systems, and 5G networks has significantly heightened the demand for robust cybersecurity solutions. Traditional network architectures face challenges in scalability, security, and manageability, leading to the adoption of software-defined networking (SDN) as a flexible and secure alternative. However, the dynamic nature of tra!c patterns in SDN necessitates frequent reconfiguration, highlighting the importance of autonomous decision-making tools. Reinforcement learning (RL), and its deep learning extension, deep reinforcement learning (DRL), have proven e”ective in addressing these challenges. Nevertheless, DRL algorithms are susceptible to adversarial attacks, underscoring the need for research into more resilient algorithms, such as NEC2DQN and DDQN, for both defensive and o”ensive applications in SDN environments. The evolution of network security has shifted from rule-based systems to autonomous solutions, with Deep Reinforcement Learning (DRL) emerging as a leading approach. Algorithms such as Deep Q-Network (DQN), Double Deep Q-Network (DDQN), and Neural Episodic Control to Deep Q-Network (NEC2DQN) have advanced the field, but their adaptability has introduced vulnerabilities to adversarial attacks, necessitating ongoing testing and improvement to ensure robust, adaptive models. | |
| dc.identifier.uri | https://hdl.handle.net/10566/20971 | |
| dc.language.iso | en | |
| dc.publisher | University of the Western Cape | |
| dc.subject | Software Defined Networks (SDN) | |
| dc.subject | Cybersecurity | |
| dc.subject | Deep Reinforcement Learning (DRL) | |
| dc.subject | Double Deep Q-Network (DDQN) | |
| dc.subject | Neural Episodic Control to Deep Q-Network (NEC2DQN) | |
| dc.title | Adversarial deep reinforcement learning for autonomous cyber defense in software defined networks | |
| dc.type | Thesis |