ABSTRACT
This paper proposes RFC-MedDP, a framework extending Resilient Federated Chain to medical imaging with differential privacy guarantees, addressing challenges in adversarial attacks and privacy in healthcare AI.
PAPER · PDF
Loading PDF…
Key findings
RFC-MedDP integrates medical-aware robust aggregation rules and adaptive differential privacy mechanisms.
The framework provides provable (ϵ, δ)-differential privacy guarantees for clinical deployment.
Theoretical analysis of privacy-utility trade-offs and comprehensive experiments using medical imaging benchmarks are conducted.
Limitations & open questions
The framework's scalability in a larger number of participating institutions is not addressed.
The impact of high-dimensional medical imaging data on the privacy-utility trade-off requires further study.