PoSFed: stake-aware validator selection and update filtering for blockchain-coordinated healthcare federated learning
Year: 2026
Type: article
Abstract: Federated learning enables collaborative model training without centralizing sensitive patient data, but reliance on a single coordinating server may introduce service-disruption and update-integrity risks. We present PoSFed, a blockchain-coordinated federated learning framework that combines PoS-inspired validator selection, validator-side F1-score-based update filtering, and contribution-aware reward allocation. PoSFed was evaluated using the Pima Indian Diabetes, Wisconsin Breast Cancer, and COVIDx CXR-4 datasets in controlled single-server simulations with 5, 10, and 20 clients under IID, non-IID, and adversarial settings. When 50% of the simulated participants submitted malicious updates, PoSFed achieved an accuracy of 86.7 ± 1.4% and an F1-score of 88.6 ± 1.2%, showing improved robustness relative to the evaluated conventional federated optimization baselines. These results provide preliminary evidence of the feasibility of stake-aware update validation in the evaluated simulation. Validation on a concrete permissioned blockchain and geographically distributed healthcare networks remains necessary.
Source: Scientific Reports
Authors: Cheng-Shun Lee, Jenn-Jier James Lien
Institutions: R.O.C Military Academy, National Cheng Kung University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7221029516