Analyzing Federated Learning as a novel approach for enhancing security and privacy in the Internet of Things (IoT)

Traditional security measures are insufficient to meet the security challenges in Internet of Things (IoT). Therefore, Artificial Intelligence (AI), particularly Machine Learning (ML), has become a reassuring solution. However, the untrustworthy network and diverse data of IoT devices amplify security concerns, highlighting the ongoing need for advanced & adaptive security measures. In this experimental study, a Federated Learning (FL) way to address issues in IoT Apps. Our method involves nine nodes, each of which trains local model. The weights in these local models are then aggregated to form a main model that can detect intrusions in network. We tested model performance using N-BaIoT & a custom dataset created in Contiki-OS on RPL protocol. Significant improvements were observed in accuracy, precision, recall, f1-score, and AUC compared to central ML approaches, with up to 99.9% enhancements. The proposed FL method showcases its efficacy in improving detection of network intrusions, marking a substantial advancement in securing the intricate IoT environment. The application of FL to address IoT security issues presents a promising route for future research & development in IoT security.