Peer-reviewed veterinary case report
How much of a face needs covering to prevent AI ID
By Cho C et al.·2026·Vanderbilt University, United States·View original on Europe PMC →
PetCaseFinder translated the abstract of this peer-reviewed paper into plain English so pet owners can read it. We do not publish original research — every detail traces back to the citation above. How we work →
Original publication title: How much of a face is a face: exploring reidentification potential with generative AI.
Plain-English summary
This study looked at how well artificial intelligence (AI) can recreate faces from photos where some features are covered up, like just the eyes. Researchers tested 10,000 facial images and found that if only the eyes or another single feature were hidden, the AI could still identify the person very accurately. However, when all major facial features, like the eyes, nose, and mouth, were covered, the AI's ability to recognize the face dropped significantly. The findings suggest that simply covering parts of a face may not be enough to protect someone's identity in clinical photos, highlighting a risk with current methods of deidentification.
Abstract
<h4>Purpose</h4>Clinical photographs play an integral role across medical fields. Since the mid-20th century, deidentification has consisted of black bars covering specific facial features, typically the eyes alone. Although increasingly questioned, this practice persists in clinical and academic settings.<h4>Approach</h4>A barrier to standardized deidentification guideline development is the unknown risk of artificial intelligence (AI) to reconstruct faces from partially obscured photos. We evaluate the ability of generative AI to reconstruct 10,000 facial images in the Synthetic Faces High Quality dataset across 14 regional masking strategies.<h4>Results</h4>Covering the eyes or any other single facial feature resulted in highly identifiable reconstructions, demonstrated by low face mesh distortion (0.14 to 0.18 relative to whole-face masking; absolute total face mesh distortion 8.34 to 10.19) and high structural similarity index to the original face (1.24 to 1.25 relative to whole-face masking; absolute SSIM 0.91 to 0.92). An open-source face verification model using Dlib was able to match 97.98% to 99.93% of these reconstructed images with the original image prior to single feature masking. Removing all major facial features (eyebrows, eyes, nose, and mouth) resulted in a threefold reduction in face verification rates compared with eyes alone, from 98.87% (95% CI [98.63%, 99.07%]) to 33.93% (95% CI [32.95%, 34.94%]).<h4>Conclusions</h4>We provide quantitative metrics of the reidentification risk that modern generative AI technology poses for partially obscured facial images.
Find similar cases for your pet
PetCaseFinder finds other peer-reviewed reports of pets with the same symptoms, plus a plain-English summary of what was tried across them.
Search related cases →Original publication on Europe PMC: https://europepmc.org/article/MED/41700218