Privacy-Preserving Monitoring And Classification Of On-Screen Activities In E-Learning Using Federated Learning

Abstract: In e-learning, tracking and classification of on-screen endeavors are fundamental for identifying learner engagement and optimizing content delivery. However, traditional methods often compromise user privacy by centralizing sensitive data. In classify to improve privacy preservation, this research suggests a novel method for tracking and classifying on-screen activities using Federated Learning (FL). Our method allows data to remain decentralized on users' devices while leveraging aggregated models for analysis. We evaluate the performance of the FL-based system against traditional centralized methods, highlighting improvements in both privacy and accuracy.