ABSTRACT The recent rapid evolution of the Internet of Medical Things (IoMT) technologies has revolutionized the process of delivering healthcare, providing the opportunity of the remote diagnostics, constant monitoring of the patients, and allowing to make clinical decisions based on the data. However, there are serious concerns associated with such developments, including the protection of very sensitive patient data, integrity of the model when trained in team‐based environments, and scalability in the resource limited and heterogeneous environments. Traditional cryptography solutions, for example, lattice‐based, polynomially based, or feature‐based cryptography, are typically linked to high key generation, communication latency and non‐scalability making them infeasible in a large‐scale IoMT application. In order to resolve these concerns, we suggest MED‐SECURE, a cryptographic‐by‐design architecture that incorporates hybrid keyless encryption, stochastic obfuscation, edge‐sensitive verification into a federated training pipeline at the device, edge and cloud layers. Unlike in the traditional key‐pair model, MED‐SECURE does not rely on the centralized key exchange instead of lightweight key exchange with session‐based temporal encoding. Extensive experiments on three heterogeneous IoMT healthcare datasets show that MED‐SECURE consistently outperforms established security frameworks such as CRYSTALS‐Kyber, QRCF, AFCP, EBPPA, and lattice‐based cryptography. By eliminating static key generation through hybrid keyless encryption, MED‐SECURE reduces initialization latency by nearly 90% (4–6 ms vs. 42–61 ms). Encryption and decryption times are reduced by 20%–48% and 24%–51%, respectively. In addition, MED‐SECURE lowers end‐to‐end communication overhead (measured in transmission latency) by up to 38% and reduces cloud‐side memory consumption during aggregation by 30%–40% under increasing transaction loads.
Dhinakaran et al. (2026) studied this question.