Authorization

 

Journal

Volume 25, No.2, Special Issue on Artificial Intelligence and Computational Methods for the Medical Cyber-Physical Systems

 
Explainable TinyML Framework for Secure and Trustworthy Federated Medical Cyber-Physical Systems (pp.332-348)
M.A. Alohali1, A.K. Samha2, A. Alshuhail 3, H. Alqahtani4, N.A. Alghanmi5, A.E. Yahya6,*
https://doi.org/10.30546/1683-6154.25.2.2026.332


1Department of Information Systems, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University,
  P.O. Box 84428, Riyadh 11671, Saudi Arabia

2Department of Management Information Systems, College of Business Administration, King Saud University, Riyadh 28095,
  Saudi Arabia

3Department of Information Systems, College of Computer Sciences and Information Technology, King Faisal University,
  Saudi Arabia

4Department of Information Systems, College of Computer Science, Center of Artificial Intelligence, King Khalid University, Abha,
  Saudi Arabia

5Department of Information Technology, Faculty of Computing and Information Technology, King Abdulaziz University, Rabigh 25732,
  Saudi Arabia

6Department of Information Technology, College of Computing and Information Technology, Northern Border University, Arar,
  Saudi Arabia

∗Corresponding author: e-mail: Abdulsamad.qasem@nbu.edu.sa




Abstract.
The paper proposes a TinyML-enabled Federated Learning (FL) architecture of Cyber-Physical Systems (MCPS), in which resource-constrained devices train local models on limited resources and transmit encrypted updates to facilitate secure aggregation. The structure incorporates both local and global explainability in order to instill more confidence in clinicians. A vital-sign classification task of synthetic features compares two models, that is a tiny linear classifier and a small neural network. Findings indicate that the linear model converges faster, has a smaller communication burden and is more interpretable, whereas the neural model achieves comparable performance with additional rounds of training. The proposed solution brings into the limelight the efficiency of explainable AI, edge-centric AI, and privacy-preserving AI in the next-generation healthcare system.


Keywords:
TinyML, Federated Learning, Explainable AI, Medical IoT, Cyber-Physical Systems, Trust, Privacy, Security, Edge Intelligence, Healthcare AI.

 

AMS Subject Classification: 68T07, 68T09, 68Q32, 94A60.



COIA 2027

7-9 July 2027
Baku, Azerbaijan

Search