IOT edge computing layer modification based cyber-attack detection using Federated-Active Learning

The Internet of Things and its practical uses are becoming more widespread as the number of connected devices increases, but it always carries a risk to network security. Therefore, it is vital for an IoT network design to rapidly and accurately identify potential attackers. While many proposed solutions focus on secure IoT algorithms, little attention has been given to reducing complexity. To address this gap, this paper proposes an IOT edge computing layer modification based cyber-attack detection edge-cloud architecture that enables quick response by detecting attacks at the Intelligent Buffalo based Secure Edge-enabled Computing layer near their source, offering versatility while decreasing the Cloud’s workload. Additionally, F-AL, a low-complexity multi-attack detection model for deployment at the edge zone, leveraging high accuracy federated active learning approaches, is introduced. The performance evaluation is conducted using the latest BoT-IoT dataset against other Machine Learning and Deep Learning methods which demonstrate that FAL outperforms SVM, FLAD, DL, DNN in terms of accuracy.