Gaze3P: Gaze-Based Prediction of User-Perceived Privacy

2025-07-01

AI Summary

This study addresses the subjectivity of user privacy perception by proposing an implicit, dynamic quantification method based on eye-tracking behavior. Methodologically, we construct Gaze3P—the first large-scale eye-tracking dataset specifically designed for privacy perception—and develop a mapping model from eye-movement trajectories to privacy sensitivity. Crucially, we establish, for the first time, a closed-loop optimization framework integrating this model with differential privacy (DP), wherein predicted privacy sensitivity guides personalized configuration of DP noise parameters. Our contributions are threefold: (1) introducing a novel eye-tracking–based paradigm for representing privacy perception; (2) publicly releasing the open-source Gaze3P dataset; and (3) empirically demonstrating that our approach significantly improves the trade-off between user experience consistency and privacy protection efficacy under DP. Both prediction accuracy and DP parameter optimization performance achieve state-of-the-art results.

Technology Category

• Machine Learning: Privacy
• Computer Vision: Bias, Fairness & Privacy
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Security and Privacy: Data transparency and provenance
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Privacy is a highly subjective concept and perceived variably by different individuals. Previous research on quantifying user-perceived privacy has primarily relied on questionnaires. Furthermore, applying user-perceived privacy to optimise the parameters of privacy-preserving techniques (PPT) remains insufficiently explored. To address these limitations, we introduce Gaze3P -- the first dataset specifically designed to facilitate systematic investigations into user-perceived privacy. Our dataset comprises gaze data from 100 participants and 1,000 stimuli, encompassing a range of private and safe attributes. With Gaze3P, we train a machine learning model to implicitly and dynamically predict perceived privacy from human eye gaze. Through comprehensive experiments, we show that the resulting models achieve high accuracy. Finally, we illustrate how predicted privacy can be used to optimise the parameters of differentially private mechanisms, thereby enhancing their alignment with user expectations.

Problem

• Predict user-perceived privacy using gaze data
• Optimize privacy-preserving techniques with gaze predictions
• Create dataset for systematic privacy perception research

Innovation

• Gaze data predicts user-perceived privacy dynamically
• Machine learning model achieves high prediction accuracy
• Optimizes differentially private mechanisms via gaze insights