Trustworthy federated learning model for the internet of robotic things
Sultan Basudan, Abdulrahman Mohammed Ahmed Alamer · Enterprise Information Systems · 2024
Federated learning (FL) has become a viable concept in the Internet of Robotic Things (IoRT) by allowing local gradients to be shared and used to train a global model without disclosing raw data. However, local data privacy leaking is a threat posed by the shared gradients. In this paper, we propose a novel FL model for resource-constrained IoRT devices to improve learning models by using local resources and client activity observation. The proposed model helps protect data privacy, reduce communication overhead and open an environment for distributed learning models. The experimental assessments executed on real-world datasets with high practical efficiency.