What question did this study set out to answer?

This research aims to address interoperability challenges in healthcare IoMT systems during the migration to post-quantum cryptography.

May 21, 2026Scientific Reports0 citationsOpen Access

HybridTrust: on-device federated learning with crypto-agile security for legacy and quantum-safe medical devices

UKUmar Hayat KhanAbdul Wali Khan University Mardan RKRahim KhanAbdul Wali Khan University Mardan SCSamia Allaoua ChellougPrincess Nourah bint Abdulrahman University

Puntos clave

This research aims to address interoperability challenges in healthcare IoMT systems during the migration to post-quantum cryptography.
Developed HybridTrust, a federated TinyML framework for ICU monitoring devices.
Integrated dynamic crypto-negotiation for selecting cryptographic schemes among ECC, ML-KEM-512, and hybrid operations.
Evaluated with a synthetic ICU dataset across five simulated hospitals in an OMNeT++ environment.
Achieved an average anomaly detection accuracy of 90.5%.
Introduced quantum-safe cryptographic overhead of only 0.09-0.43 ms per operation.
Facilitated secure transitions and interoperability without significant latency or overhead.

Resumen

The process of migration of IoMT systems in healthcare into post-quantum cryptographic systems is expected to be a gradual one. Here, existing ECC-based devices alongside newly developed quantum-resistant devices will be operating within the same healthcare ecosystem. However, there exists an important interoperability challenge posed by the need for collaboration within the domain of edge intelligence in critical life situations like ICU monitoring. This paper presents HybridTrust, an on-device federated TinyML framework designed to enable secure collaboration among legacy, post-quantum, and hybrid IoMT devices during the post-quantum cryptography migration period. Unlike wearable IoMT systems, HybridTrust targets mains-powered ICU monitoring devices, where cryptographic flexibility, clinical fidelity, and security can be prioritized without strict battery constraints. The novelty of HybridTrust lies in its crypto-agile and deployment-oriented design rather than in proposing a new machine-learning algorithm. The framework integrates a dynamic crypto-negotiation protocol that selects the strongest mutually supported scheme among ECC, ML-KEM-512, and hybrid ECC-ML-KEM operation; a compact 931-byte hybrid certificate format suitable for ESP32-class microcontrollers; a regulatory compliance scorecard mapping cryptographic choices to HIPAA, GDPR, FDA, and NIST expectations; and a tree-concatenation-based model fusion strategy for maintaining clinical utility under non-IID hospital data distributions. HybridTrust is evaluated using a clinician-validated synthetic ICU dataset distributed across five simulated hospitals in an OMNeT++ environment. The framework achieves an average anomaly detection accuracy of 90.5% while introducing only 0.09-0.43 ms quantum-safe cryptographic overhead per operation. These results indicate that secure post-quantum transition, legacy interoperability, and real-time ICU monitoring can be jointly supported without imposing prohibitive latency or deployment overhead. HybridTrust therefore provides a practical pathway for backward-compatible, crypto-agile, and quantum-resilient IoMT deployment in critical-care environments.

Preguntar a la IA

Me gusta

Guardar

Ver artículo completo