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

Federated Learning and Artificial Intelligence (AI) are two most intriguing and leading technologies in the intelligent healthcare business. Data must be collected, stored and analyzed from various companies. Patient data processing, particularly in the medical industry, promises substantial breakthroughs in customized health care. For example, different hospitals preserve electronic health records (EHR) for different patient populations, which are difficult to communicate between hospitals due to their concern. As electronic health record (EHR) systems have developed more widespread, privacy and security concerns have increased significance. Sharing and analyzing sensitive health data across multiple organizations and stakeholders requires the application of sophisticated privacy-protection strategies. This study offerings a Secure Federated Transfer Learning (SFTL) structure with an improved system to address privacy and security issues in smart EHR systems. While preserving the privacy of individual patient records, the SFTL framework enables various healthcare providers to interact and collaboratively train machine learning (ML) models using their dispersed EHR data. SMPC enhances privacy guarantees by securely calculating statistical metrics and model parameters across multiple participants without disclosing their contributions. Based on a real-world smart EHR dataset, we evaluate the proposed SFTL -SMPC architecture and compare its performance to existing privacy-preserving techniques. The results demonstrate that our strategy, SFTL-SMPC, balances data privacy and model precision, outperforming traditional federated learning approaches while maintaining patient data security. In addition, we conduct a comprehensive security analysis to validate that our SFTL-SMPC implementation is resilient against various attack scenarios.