Privacy Risks of Robot Vision: A User Study on Image Modalities and Resolution

TL;DR

This study addresses privacy risks associated with visual data collected by mobile service robots operating in private spaces. It uses a user study with 62 participants to compare RGB, depth, and semantic segmentation modalities and to identify resolution thresholds for privacy preservation. Key findings show RGB is perceived as the most privacy-invasive, while depth and semantic segmentation are viewed as privacy-friendly; participants favor capturing low-resolution data at the source, with $32 \times 32$ as a basic privacy threshold and $16 \times 16$ for stronger protection. These insights inform privacy-by-design guidelines for robot vision systems, emphasizing early data minimization, modality selection, and clear privacy thresholds to balance performance and user comfort.

Abstract

User privacy is a crucial concern in robotic applications, especially when mobile service robots are deployed in personal or sensitive environments. However, many robotic downstream tasks require the use of cameras, which may raise privacy risks. To better understand user perceptions of privacy in relation to visual data, we conducted a user study investigating how different image modalities and image resolutions affect users' privacy concerns. The results show that depth images are broadly viewed as privacy-safe, and a similarly high proportion of respondents feel the same about semantic segmentation images. Additionally, the majority of participants consider 32*32 resolution RGB images to be almost sufficiently privacy-preserving, while most believe that 16*16 resolution can fully guarantee privacy protection.