PQ-TA-ASE-FGNN: Post-Quantum Trust-Aware Adaptive Searchable Encryption Using Heterogeneous Federated Graph Neural Networks for Quantum-Resilient Healthcare Insider Threat Detection

Abstract

Healthcare systems increasingly depend on federated multi-institutional architectures in which IoMT devices, electronic health records, and access control systems exchange sensitive model parameters across institutional boundaries. Classical cryptographic primitives, specifically ECDH key encapsulation and ECDSA digital signatures, are broken by Shor’s algorithm on a Cryptographically Relevant Quantum Computer (CRQC), enabling Harvest-Now-Decrypt-Later (HNDL) attacks on archived federated gradient transmissions. This paper presents PQ-TA-ASE-FGNN, extending the prior TA-ASE-FGNN system with four contributions: (1) trust-conditioned ML-KEM-768 session key encapsulation, (2) Merkle-batched ML-DSA-65 blockchain signing reducing storage by 93%, (3) a Crypto-Agile Interface Layer, and (4) a formal HNDL threat model. Evaluated on four datasets (synthetic IoT healthcare with 100,000 nodes, MIMIC-III, CICIoMT2024, and CERT Insider Threat v6.2), the framework achieves F1-scores of 0.93, 0.89, 0.88, and 0.91 respectively while maintaining the 2-second revocation SLA at 1.84 seconds under concurrent load