Verifiable Blockchain-Empowered Federated Learning for Secure Data Sharing in the Internet of Medical Things
The increasing amount of data produced by medical devices in the Internet of Medical Things creates new opportunities to improve the quality of medical services through data sharing in medical institutions. However, data sharing risks both external attacks and unauthorized internal personnel access...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
American Association for the Advancement of Science (AAAS)
2025-01-01
|
| Series: | Intelligent Computing |
| Online Access: | https://spj.science.org/doi/10.34133/icomputing.0132 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The increasing amount of data produced by medical devices in the Internet of Medical Things creates new opportunities to improve the quality of medical services through data sharing in medical institutions. However, data sharing risks both external attacks and unauthorized internal personnel access to private data. Unauthorized access can lead to severe consequences beyond financial loss for the providers. This paper proposes a verifiable blockchain-empowered federated learning framework to address privacy and aggregation verification challenges in data sharing. An efficient verification mechanism is developed using a validation committee to filter out inadequate training results and maliciously uploaded gradients, thereby preventing Byzantine attacks and improving the security and efficiency of global model aggregation. Additionally, a contribution-based weighted aggregation scheme is proposed, which assigns coefficients based on participant contributions, providing a more thoughtful approach than the traditional federated averaging algorithm. To tackle node selection for global model aggregation and federated learning updates, a proof of contribution consensus mechanism is introduced. Theoretical assessments and performance metrics indicate that the proposed method is both efficient and secure, considerably enhancing the accuracy of the global model. |
|---|---|
| ISSN: | 2771-5892 |