Architecture for Secure and Real-Time Patient Monitoring with Wearable and Implantable Devices (pp. 227-244)
M. Maashi1, *, H.A. Mengash2, A. Almutairi3, J.Khan4, T. Althobaiti5, A. Alsayat6, R. Alsini7
https://doi.org/10.30546/1683-6154.25.2.2026.227 1Department of Software Engineering, College of Computer and Information Sciences, King Saud University, Po Box 103786,
Riyadh 11543, Saudi Arabia
2Department of Information Systems, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University,
P.O. Box 84428, Riyadh 11671, Saudi Arabia, e-mail: hamengash@pnu.edu.sa
3Department of Information Systems, College of Computer Science and Information Technology, King Faisal University,
P.O. Box 400, Al-Ahsa 31982, Saudi Arabia, e-mail: akalmutairi@kfu.edu.sa
4Department of Computer Science, Applied College at Mahayil, King Khalid University, Saudi Arabia, e-mail: jhkhan@kku.edu.sa
5Department of Computer Science, Faculty of Science, Northern Border University, Arar 73222, Saudi Arabia,
e-mail: Turke.althobaiti@nbu.edu.sa
6Department of Computer Science, College of Computer and Information Sciences, Jouf University, Sakaka 72341,
Saudi Arabia, e-mail: asayat@ju.edu.sa
7Department of Information Systems, Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah,
Saudi Arabia, e-mail: raalsinie@kau.edu.sa
∗Corresponding author: e-mail: mmaashi@ksu.edu.sa
Abstract. A Federated Edge Learning (FEL) architecture, described in this paper, was developed to facilitate patient monitoring as a scalable, secure and real-time feature. The proposed FEL framework integrates heterogeneous biomedical IoT devices, such as ECG patches, continuous glucose monitor, neurostimulators, and implantable cardiac devices, which can communicate with localized edge gateways and can engage in the federated model training with preprocessing performed on the devices. A hybrid CNN-LSTM classifier is applied to multimodal biosignal classification and considers the constraints of wearable and implantable devices. Privacy preserving technologies such as homomorphic encryption and secure aggregation are used to protect model updates to gradient inversion attacks and model poisoning attacks, and differential privacy enhances data confidentiality. The simulation was based on synthetic multimodal biomedical data (ECG, glucose, and temperature signals) through a comprehensive MATLAB-based simulation to measure FEL performance alongside centralized learning (CL) and cloud-assisted edge learning (CAEL) baselines. Experiments show that FEL outperforms CAEL and CL in average classification accuracy (96.2% versus 91.8% versus 89.7%), yet still performs 36% faster in terms of end-to-end latency and 29% cheaper in terms of communication cost due to on-device feature extraction. The proposed framework further showed adaptive and more trustworthy security with adversarial and privacy-constrained settings, further proving its feasibility in next-generation digital healthcare systems of high priority: Low latency, high accuracy, and privacy compliance.
Keywords: Federated Edge Learning, Wearable Devices, Implantable Devices, CNN-LSTM, Secure Aggregation, Homomorphic Encryption, Differential Privacy, Internet of Medical Things (IoMT), Real-Time Patient Monitoring, Edge Computing.
AMS Subject Classification: 68T07, 68M10, 68M14, 92C50, 94A60.