Screen Activity Monitoring Using Federated Learning

Nowadays, students are distracted by different entertainment websites and engage in non-academic activities while they are supposed to be online studying. To address this issue, an architecture safeguarding privacy is developed to assess whether students are effectively managing their computer usage. A dataset with over 3,500 screenshots capturing diverse student computer interactions has been used to categorize them into different classes. Multiple pre-trained models were utilized for this task. FedInceptionV3 is a variant of the InceptionV3 model which is known for its high accuracy of 98.25% in classifying images. This is an effective approach to accurately classify and monitor student activities. It handles user data with care and uses privacy-preserving measures to respect user privacy.