The rapid growth in the scale and complexity of modern networks has significantly increased the challenges associated with their management, maintenance, and optimization. Software-defined networking (SDN) has emerged as a transformative paradigm, offering centralized control, a global network perspective, and programmable traffic handling to address these issues effectively. Despite its advantages, the centralized architecture of SDN introduces critical security vulnerabilities, particularly to cyber threats such as Denial-of-Service (DoS) attacks. Conspicuously, a range of security strategies has been proposed, including statistical, threshold-based, and machine learning (ML)-driven techniques. However, Deep Learning (DL) models have demonstrated superior performance in detecting and mitigating attacks due to their ability to learn complex patterns within network traffic data. This survey presents a comprehensive analysis of recent advancements in the application of DL methods for SDN security. It systematically categorizes attack types targeting SDN, reviews DL-based detection and mitigation approaches, and evaluates the public datasets employed for model training, highlighting their benefits and limitations. The paper concludes by identifying key challenges and outlining promising directions for future research to enhance the effectiveness and adaptability of DL solutions in securing SDN infrastructures.
Shyryn et al. (Fri,) studied this question.