Research on medical data security sharing scheme based on homomorphic encryption.

Area of Science:

• Artificial Intelligence in Medicine
• Medical Data Security
• Federated Learning

Background:

• AI-assisted technologies enhance disease prediction and diagnosis accuracy in medicine.
• Data security concerns impede crucial data sharing among healthcare institutions.
• Need for secure methods to leverage big data for medical advancements.

Purpose of the Study:

• To develop a secure medical data sharing scheme for collaborative analysis.
• To implement a federated learning architecture protecting training parameters via homomorphic encryption.
• To enable AI-driven disease prediction without compromising patient privacy.

Main Methods:

• Developed a Client/Server (C/S) communication mode for medical data sharing.
• Utilized federated learning with the Paillier algorithm (additive homomorphism) for parameter encryption.
• Employed stochastic gradient descent for model training and parameter updates on clients.

Main Results:

• The proposed scheme enables secure data sharing and collaborative model training.
• Achieved accurate disease prediction while ensuring patient data privacy.
• Model prediction accuracy is influenced by global training rounds, learning rate, batch size, and privacy budget.

Conclusions:

• The developed scheme effectively balances data sharing needs with robust privacy protection.
• Demonstrated the feasibility of using federated learning and homomorphic encryption in medical AI.
• The system shows promising performance for secure, AI-assisted medical diagnostics.