Hybrid Intrusion Detection System for Software-Defined Networks

Software-Defined Networking, as a relatively recent networking paradigm, offers centralized infrastructure management, flexibility and high programmability. However, it also creates particular security risks due to being exposed to external threats. To address these challenges, numerous methods have been developed and applied over the past few years. This study proposes a hybrid Intrusion Detection System that combines signature-based analysis with deep learning-based anomaly detection. In this architecture, a signature module quickly filters known attack patterns, while remaining traffic is analyzed by an autoencoder and a supervised deep neural network classifier. The final decision is based on rule-based prioritization of the outputs from both models, improving the reliability and robustness of detection.