Secure federated transfer learning with enhanced secure multiparty computation for privacy preserving smart EHR...

Area of Science:

• Artificial Intelligence
• Machine Learning
• Healthcare Informatics

Background:

• Electronic Health Records (EHR) are crucial for personalized healthcare but raise significant privacy and security concerns.
• Sharing sensitive patient data across institutions is challenging due to privacy requirements and data silos.
• Federated Learning (FL) and Artificial Intelligence (AI) offer potential for healthcare advancements but require robust privacy-preserving methods.

Purpose of the Study:

• To propose a Secure Federated Transfer Learning (SFTL) framework integrated with Secure Multi-Party Computation (SMPC) for smart EHR systems.
• To address the critical privacy and security challenges in collaborative analysis of distributed patient data.
• To enable healthcare providers to train machine learning models on dispersed EHR data without compromising individual patient privacy.

Main Methods:

• Development of a Secure Federated Transfer Learning (SFTL) architecture.
• Integration of Secure Multi-Party Computation (SMPC) for privacy-preserving statistical calculations and model parameter aggregation.
• Evaluation of the SFTL-SMPC framework using a real-world smart EHR dataset.

Main Results:

• The SFTL-SMPC framework effectively balances data privacy and model accuracy.
• The proposed approach demonstrates superior performance compared to traditional federated learning methods.
• The SFTL-SMPC implementation shows resilience against various security threats in a comprehensive analysis.

Conclusions:

• The SFTL-SMPC framework provides a secure and effective solution for privacy-preserving collaborative learning in smart EHR systems.
• This approach facilitates the advancement of AI in healthcare by enabling secure data utilization.
• The study validates the robustness and efficacy of the SFTL-SMPC architecture for sensitive health data analysis.