Fairly Private: Investigating The Fairness of Visual Privacy Preservation Algorithms

TL;DR

This paper examines the fairness of visual privacy preservation algorithms by analyzing how well they protect privacy across different groups, revealing disparities in performance.

Contribution

It introduces a fairness analysis of privacy algorithms, highlighting unequal protection across demographic groups in facial recognition scenarios.

Findings

Privacy protection varies significantly across groups.

Current algorithms do not ensure equitable privacy preservation.

Facial recognition performance on obfuscated images is biased.

Abstract

As the privacy risks posed by camera surveillance and facial recognition have grown, so has the research into privacy preservation algorithms. Among these, visual privacy preservation algorithms attempt to impart bodily privacy to subjects in visuals by obfuscating privacy-sensitive areas. While disparate performances of facial recognition systems across phenotypes are the subject of much study, its counterpart, privacy preservation, is not commonly analysed from a fairness perspective. In this paper, the fairness of commonly used visual privacy preservation algorithms is investigated through the performances of facial recognition models on obfuscated images. Experiments on the PubFig dataset clearly show that the privacy protection provided is unequal across groups.