OcuGuard: A Privacy-Preserving Mobile Multimodal Framework for Real-Time Digital Eye Strain Mitigation

Prolonged use of digital displays, combined with widespread shifts to remote work and education, has coincided with a marked rise in myopia across children and adults. Existing screen-monitoring tools mainly provide retrospective, time-based logs and lack real-time physiological data or vision-specific diagnostics. This paper introduces OcuGuard, a privacy-centric, cross-platform mobile framework for iOS and Android that continuously monitors and mitigates digital eye strain using multimodal sensor fusion. With real-time computer vision processed on a secure backend, OcuGuard tracks gaze, blink frequency and dynamics, viewing distance, and ambient light. These data streams feed a machine learning regression model, calibrated with isotonic regression, to compute an individualized daily eye strain index. Users see this index through an interpretable Eye Nutrition Label, visually modeled on the format of the U.S. Food and Drug Administration (FDA) Nutrition Facts panel (not an FDA-endorsed or regulatory label). We also design and implement closed-loop, automated interventions—such as context-aware display brightness control, micro-blink reminders, and gamified eye exercises—while transmitting only short-lived, encrypted video frames to a backend for feature extraction (frames are discarded immediately after processing and never persisted) and keeping derived scores and long-term storage on the user’s device by default, sharing only aggregated summary reports with an explicitly configured guardian account (Section 3.4) to protect privacy. The pipeline is architected to support alignment with the principles of the Children’s Online Privacy Protection Act (COPPA) and the Health Insurance Portability and Accountability Act (HIPAA).