Blockchain and homomorphic encryption based privacy-preserving model aggregation for medical images

Area of Science:

• Artificial Intelligence in Healthcare
• Medical Image Analysis
• Blockchain Technology

Background:

• Traditional AI in healthcare requires centralized data, posing privacy risks.
• Sharing sensitive patient data for AI model training is a significant challenge.
• Existing methods lack robust security for collaborative medical data analysis.

Purpose of the Study:

• To propose a novel blockchain-based federated learning framework for secure COVID-19 image analysis.
• To enhance data privacy and security in collaborative medical AI model training.
• To develop a decentralized approach for sharing encrypted federated models.

Main Methods:

• Utilized a novel capsule network for COVID-19 image segmentation and classification.
• Implemented a homomorphic encryption scheme to secure model gradients.
• Developed a blockchain-based federated learning algorithm for decentralized model sharing.

Main Results:

• The proposed framework successfully trains local models using encrypted gradients.
• Blockchain ensures decentralization, eliminating reliance on central servers.
• Experimental results demonstrate the superior performance and security of the proposed scheme.

Conclusions:

• The integration of blockchain and federated learning offers a new paradigm for secure medical image data sharing.
• The framework effectively preserves data privacy while enabling collaborative AI model development.
• This approach addresses critical security and privacy concerns in AI-driven healthcare.