Edge-based Federated Learning to Enhance the Security of IoT

Abstract

The integration of Artificial Intelligence (AI) and Internet of Things (IoT) has enabled network components to make independent decisions, but this capability also poses a risk of malicious attacks in unsupervised and trustless environments. To overcome this issue, the article proposes a distributed collaborative detection approach that utilizes edge nodes as voters to monitor the training process, tackle PA, and improve the accuracy of the global model. The proposed approach is evaluated using the UNSW-NB15 dataset in both IID and non-IID scenarios, and the results demonstrate the effectiveness of the approach in improving the accuracy of FL even in the presence of PA.