Privacy Preserving On-Screen Activity Tracking and Classification in E-Learning using Federated Learning

In the rapidly evolving landscape of remote and online learning, the ability to monitor and assess students' engagement and productivity has become increasingly vital. This paper presents a pioneering approach to address this challenge through "Privacy-Preserving On-Screen Activity Tracking and Classification in E-Learning Using Federated Learning." Our innovative solution combines the benefits of real-time user monitoring with stringent privacy protection, aiming to distinguish whether students are effectively utilizing their time for knowledge development or, regrettably, wasting it. E-learning platforms have gained immense popularity, especially in light of global events necessitating remote education. However, ensuring that students are actively engaged and focused during online sessions remains a significant concern. Our approach leverages Federated Learning, a decentralized machine learning paradigm, to safeguard user privacy while accurately classifying on-screen activities. By employing this technique, we address the dual challenge of preserving user privacy and providing valuable insights into the efficiency of online learning. Federated Learning empowers our system to train machine learning models collaboratively across multiple user devices, eliminating the need to centralize sensitive data on a single server. This ensures that individual user data, including screen recordings and interactions, remains on their devices, preserving user privacy and complying with stringent data protection regulations such as GDPR and HIPAA. This decentralized approach sets our solution apart from traditional methods that risk privacy breaches and data security concerns. Our system's classification capabilities are built upon deep learning models that analyze on-screen activities in real-time. By considering various factors such as user interactions, screen content, and time management patterns, our system can effectively differentiate between productive knowledge development and unproductive behavior, such as distractions or disengagement. This information empowers educators and institutions to take timely corrective actions, thereby enhancing the overall effectiveness of online education. "Privacy-Preserving On-Screen Activity Tracking and Classification in E-Learning Using Federated Learning" represents a groundbreaking contribution to the field of e-learning monitoring. By integrating Federated Learning with advanced classification models, we strike a delicate balance between user privacy and educational effectiveness. This innovative approach enables educators to make informed decisions, fosters student engagement, and ultimately ensures that time spent in e-learning environments is optimized for knowledge development rather than wasted. As the landscape of education continues to evolve, our system provides a robust foundation for privacy-preserving, data-driven, and effective e-learning practices.

KEYWORDS: CNN, decision tree, linear discriminant analysis.