Enhancing Malicious Traffic Detection in IoT-Enabled SDN Using Convolutional Neural Network

The Internet of Things is a captivating area of research, involving the interconnection of devices for information exchange across networks. With a growing inclination towards integrating IoT into daily lifestyles, there is a surge in the number of devices and the volume of exchanged data. However, the conventional network structure faces challenges in efficiently managing these devices and data while safeguarding privacy and data integrity from malicious intent. Software-Defined Networking emerges as a modern network design, reorganizing the traditional network into three layers to enhance flexibility and efficiency, with the control plane serving as the SDN’s core. Nonetheless, the control plane is vulnerable to attacks, posing a substantial risk to IoT devices. To address this risk, an automated model can be incorporated into an intrusion detection system, leveraging a Convolutional Neural Network to accurately classify traffic types. This study specifically assesses the effectiveness of our CNN model on the InSDN dataset, maintaining all features and observations without altering the dataset size through undersampling, oversampling, or feature selection.The numerical results showed that our CNN model achieved 100% effectiveness across all evaluation metrics for multi-class(8 labels) classification, including accuracy, precision, recall, and F1-score.