Publications

カンファレンス (国際) Context-Preserving Content Replacement for Privacy Protection in Screen Sharing with Generative AI

Rimi Makita (Ochanomizu University), Kaori Ikematsu, Yuki Igarashi (Ochanomizu University)

The ACM Symposium on User Interface Software and Technology (UIST2025)

2025.9.27

As online conferencing tools become increasingly common, users face a growing risk of unintentionally disclosing private information during screen sharing. An existing system employs masking or blurring to obfuscate such information, but it remains unclear whether these methods are appropriate in terms of privacy protection and viewing experience. An investigation of visual obfuscation methods for screen content suggested that users tend to prefer those with a more natural appearance. In this study, we propose a method for replacing private screen information with visually natural, context-preserving content. Image regions are replaced using generative AI, while text and icons are replaced using curated datasets, allowing tailored obfuscation based on content type. This method introduces a novel direction for protecting privacy during screen sharing, balancing effectiveness with viewing experience.

Paper : Context-Preserving Content Replacement for Privacy Protection in Screen Sharing with Generative AI新しいタブまたはウィンドウで開く (外部サイト)