Facial recognition cameras capture wider views than users expect, raising privacy concerns
A Fudan University professor's warning that facial recognition cameras capture wider views than the on-screen indicator suggests prompted Chinese media coverage about privacy risks. Experts clarified that legitimate apps extract encrypted feature data rather than photos, but cautioned users against irregular applications that could misuse complete facial images for deepfake fraud or extortion.
What happened
A video by Fudan University professor Zhang Junping warning users not to perform facial recognition verification while bathing sparked widespread discussion in China. The video explained that while facial recognition apps display a circle or square on screen as a positioning guide, the camera actually captures the complete front-facing camera view, not just the framed area.
According to network security experts from Qianzhen Information Technology, the actual camera frame retrieved is the complete front-facing view, and backend reviewers can see the complete picture. Many users expressed concern after learning they had performed facial verification while unclothed or in compromising situations.
Reported impact
Users reported concerns that personal privacy could be leaked, particularly those who had performed facial recognition verification while bathing, using the toilet, or in states of undress. Experts warned that irregular applications or those installed through suspicious channels could steal facial information for uses including AI deepfake fraud or extortion.
Index analysis
Experts explained that legitimate applications typically do not upload raw photographs, but instead collect facial feature data such as eyebrow-eye distance and cheekbone contours to create a data package for comparison. According to the 2025 Facial Recognition Technology Application Security Management Measures effective June 1, 2025, legitimate applications must encrypt collected facial feature data and store it in secure environments to prevent data leakage.
Security experts cautioned against applications installed through suspicious QR codes or unofficial channels, which may collect complete photos and backgrounds rather than encrypted feature data. The risk exists that such data could be used for AI face-swapping fraud, illegal account registration, forgery of compromising photos, or extortion.
Governance controls
Experts advised users to follow a principle of 'no authorization unless necessary' when granting application permissions for facial recognition, carefully reviewing relevant terms to understand the purposes for which facial recognition information will be collected and used, and avoiding using facial authentication functions on unfamiliar applications.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.