INNOVATIVE NEURAL NETWORK STRATEGIES FOR DETECTING CHEATING IN VIRTUAL EXAMS

The rise of virtual exams has introduced new challenges in maintaining academic integrity, particularly in detecting collaborative cheating. In this study, we propose a novel framework that combines Graph Neural Networks (GNNs) with Federated Learning (FL) to detect collaborative cheating while preserving student privacy. Our approach models student interactions as a graph, where nodes represent students and edges represent potential cheating relationships. By training the GNN in a federated manner, we ensure that sensitive data remains on students' devices, addressing ethical and legal concerns.