FLFL: Federated Latent Factor Learning for Private Recovery of Spatio-Temporal Signals

The paper introduces Federated Latent Factor Learning (FLFL), a model designed for privacy-preserving recovery of spatio-temporal signals in wireless sensor networks (WSNs). FLFL employs a federated learning framework that allows sensors to upload only gradient information instead of raw data, and integrates spatio-temporal correlations as a regularization constraint to enhance recovery accuracy. Experimental results indicate that FLFL achieves superior performance in recovery accuracy compared to eight existing federated and non-federated models, making it a significant advancement for practitioners concerned with data privacy in sensor data recovery.