Trustworthy Artificial Intelligence -based federated architecture for symptomatic disease detection

• Published in: Neurocomputing (Amsterdam) 2024-04, Vol.579, p.127415, Article 127415
• Main Authors: López-Blanco, Raúl, Alonso, Ricardo S., Rodríguez-González, Sara, Prieto, Javier, Corchado, Juan M.
• Format: Article
• Language: English
• Subjects: COVID-19; Federated learning; Healthcare; Internet of Things; Trustworthy Artificial Intelligence
• ISSN: 0925-2312; 1872-8286
• DOI: 10.1016/j.neucom.2024.127415

The recent viral outbreaks have had a significant impact on interpersonal relationships, particularly in enclosed spaces. Detecting and preventing the transmission of diseases such as COVID-19 has become a top priority. These diseases are typically identifiable through the symptoms they cause in humans. However, the collection of personal and health data for use in Artificial Intelligence models can give rise to ethical, security, and privacy issues. Therefore, it is necessary to have architectures that maintain the principles of Trustworthy Artificial Intelligence by design. This work proposes a decentralised architecture based on Federated Learning for symptomatic disease detection using the edge computing paradigm, storing the information in the device that collected it, and the foundations of Trustworthy Artificial Intelligence. The architecture is designed to be robust, secure, transparent, and responsible while maintaining data privacy. The proposed approach can be used with medical information capture systems with different user profiles. [Display omitted] •FBAC architecture designed to provide robustness and advocate data transparency.•Device networks that capture biometric parameters while protecting privacy rights.•Distributed computation without a central node, reducing information exchange times.•Resilience scales with the number of nodes, reducing the chance of information loss.•FBAC aims for fairness during model training and data privacy in data transport.