DeepSDN: Deep Learning Based Software Defined Network Model for Cyberthreat Detection in IoT Network

The Internet of Things (IoT) presents new challenges to traditional communication models, particularly in terms of security, which are exacerbated by the rapid evolution of cyberthreats. Traditional security methods, especially those using Machine Learning models, often struggle with the limited computational resources available, making it difficult to detect attacks across the entire network. Software-defined networks (SDN) offer a solution by centralizing security policies, enabling more effective implementation and enforcement. The study investigates the SDN architecture from a security perspective. This article proposes a Deep Learning-based SDN architecture for IoT security which can significantly enhance real-time cyberthreat detection. Specifically, a secure communication channel is first designed using a blockchain-based authentication to resist well-known intruders. Second, a deep learning convergence model using an adaptive threshold scoring method that stops all local model training and allows edge models to contribute to the cloud model until a specified accuracy is achieved. To achieve low CPU usage and provide real-time services, SDN is next used as a cloud-based system security administrator to protect IoT networks from zero-day attacks by sending requests from edge devices to a cloud SDN controller. The efficiency of the proposed framework is demonstrated using simulations with two different network datasets E-IIoT and ToN-IoT against various attacks and the results are compared with similar works. The proposed model effectively detects and mitigates cyberthreats such as DDoS, Black-Sink-Worm hole, MitM, and Ransomware. It achieves high performance with 99.15% accuracy, 99.31% precision, 98.97% recall, and a 99.14% F1 score, on an average while using less CPU power for real-time IoT network protection.