An Adaptive Hierarchical Framework for Federated Learning in the Internet of Things using Energy-Aware Client Selection and Local Personalization
An Adaptive Hierarchical Framework for Federated Learning in the Internet of Things using Energy-Aware Client Selection and Local Personalization
Rezvan Mahmoudie1
1) استادیار گروه مهندسی کامپیوتر،واحد اسدآباد دانشگاه آزاد اسلامی،اسدآباد ،ایرانا
IOI Code
:
AAHD-ABBBBC
Publication :
National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
The explosive growth of Internet of Things (IoT) devices has led to the generation of a massive volume of sensitive data. Transmitting this data to cloud servers for training machine learning models poses serious challenges regarding privacy, bandwidth, and latency. Federated Learning (FL) has emerged as a promising solution by performing model training on edge devices. However, implementing FL in IoT environments faces accuracy degradation and inefficiency due to data heterogeneity, severe resource constraints (battery and processor), and communication noise. In this paper, a novel framework named EH-PFL is proposed. By utilizing a three-layer hierarchical architecture (IoT, Edge, Cloud) and a client selection mechanism based on energy and data quality scoring, the proposed method reduces communication overhead. Furthermore, to address the data heterogeneity problem, a personalization technique based on local fine-tuning is employed. Simulation results demonstrate that the proposed method increases accuracy by up to 12% and reduces energy consumption by up to 25% compared to baseline methods such as FedAvg.
Keywords : federated learning internet of things edge computing privacy preservation personalization energy management