| This research presents a Proof of Concept designed to enhance the security and monitoring of water treatment plants by integrating mobile devices (smartphones/tablets) as “secure sensors” decoupled from traditional information systems to prevent cyberattacks, using their cameras to detect critical variables such as foam and turbidity. Through the development of computer vision models (YOLO11 and convolutional neural networks) with both online and offline execution options, the study considers federated learning to enable collaborative multi-plant management. This approach allows for the training of a robust global model without sharing sensitive local data, ensuring privacy and optimizing the control of multiple facilities from a single mobile device in a cyber-secure environment. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.