Enhancing Privacy in Smart Grids and IOTs Systems by Using Federated Learning: Case Study

Abstract

To enhance privacy in smart grids (SGs) and inter-net of thing (IoT) systems, a Federated Learning (FL) frameworkis proposed for practical application. By leveraging the idea ofdecentralizing model training and keeping raw data local, theframework addresses the privacy and security challenges associ-ated with data collection on centralized servers. The frameworkachieves high accuracy (98.2% on MNIST, 85.6% on CIFAR-10) while resisting poisoning attacks and scaling efficiently byintegrating differential privacy and secure aggregation. A casestudy on energy demand forecasting confirms its real-worldapplicability. The results demonstrate the potential of FL forscalable, privacy-preserving data analysis in IoT and SGs, withfuture work focused on integrating other privacy-enhancingtechnologies such as blockchain (BC).