Study Re-evaluates Adversarial Facial Privacy Methods and Their Reported Protection Rates
A new arXiv paper argues that published face-privacy methods claiming over 90% protection against black-box recognition models rely on an evaluation protocol with two flaws, prompting a re-assessment of those results.
Read the full story at arXiv cs.CV ↗
Timeline · 1 report
- 2026-11-18 05:00 · arXiv cs.CV
Wrong Input, Wrong Threshold: Re-evaluating Adversarial Facial Privacy