FEDERATED LEARNING FOR PRIVACY-PRESERVING ONSCREEN ACTIVITY RECOGNITION IN E-LEARNING

Abstract

The digital classroom of today enables teachers to monitor the screen activity of their students, which enables them to assess their level of engagement and pinpoint areas where their learning could be enhanced. There are additional privacy concerns that have been raised by centralized behavior monitoring technology. Federated Learning (FL) is the primary focus of this article due to its secure, screen-aware approach to online learning. FL guarantees the security of its customers' personal information by training models locally on their devices and storing model updates instead of raw data. Our federated architecture ensures user privacy by precisely monitoring student actions through deep learning, which encompasses reading, viewing, and engagement. Experimental data suggests that federated learning (FL) is a promising alternative for large-scale e-learning systems, as it has a high recognition success rate and minimal privacy concerns. The findings of this study underscore the urgent need to identify methods to protect the privacy of students in the rapidly developing field of personalized digital education.