Description of Invention
This study proposes an innovative approach integrating Federated Learning (FL) with Homomorphic Encryption (HE) using the CKKS scheme for privacy-preserving diabetes diagnosis. FL decentralises data processing, keeping sensitive patient data on local devices, while CKKS-based HE ensures computations are done on encrypted data, preserving privacy during model training and aggregation. The approach achieves 92.28% accuracy after five rounds, comparable to centralised models, while maintaining data privacy. This scalable solution addresses both performance and security concerns and can be adapted to various healthcare applications, advancing privacy-preserving techniques for secure, data-driven medical innovations.