Federated Learning for AI-Assisted Education: Privacy-Preserving and Cross-Platform Collaborative Modeling in Higher Education

The rise of AI-assisted education systems in universities has introduced new opportunities for personalized learning, yet it also raises serious concerns about data privacy and platform fragmentation. Traditional centralized training approaches are incompatible with the sensitive nature of student data and the heterogeneous infrastructure across academic institutions. In this study, we propose a federated learning framework tailored for higher education environments to enable privacy-preserving AI model training across decentralized platforms. Our framework supports adaptive aggregation, differential privacy, and knowledge distillation to address both data security and model heterogeneity. Experimental results on multi-campus datasets show that the proposed system achieves comparable accuracy to centralized baselines while significantly reducing privacy leakage and communication overhead. This research provides a scalable and secure foundation for AI-driven educational analytics, fostering cross-institutional collaboration without compromising individual data confidentiality.