Designing Privacy-Protecting System with Visual Masking Based on Investigation of Privacy Concerns in Virtual Screen Sharing Environments

Abstract

Virtual meeting tools, such as Zoom, can sometimes lead to inadvertently sharing private information through screen sharing features. We conducted fundamental investigations on what concerns and strategies users have for protecting their privacy. The results indicate that while most users take actions to protect their privacy before or during meetings, a significant number of users also reported experiences with inadvertent information sharing. We also found that users avoid sharing not only personal data, such as usernames, but also information that could reveal their interests and activities, such as browsing histories or personalized recommendations. Thus, we propose a system that automatically occludes areas that users do not want to share to facilitate the management of privacy during screen sharing. We conducted a data collection experiment to construct a deep learning model to detect the areas that users do not want to share with others. We implemented our system by using this model to protect information in real time during screen sharing in virtual meeting tools.