Fairness Properties of Face Recognition and Obfuscation Systems

TL;DR

This paper investigates the demographic fairness of face obfuscation systems that generate perturbed images to evade recognition, revealing biases in embedding networks that impact minority groups.

Contribution

It provides an analytical and empirical analysis showing demographic disparities in face obfuscation effectiveness due to clustering in embedding networks.

Findings

Embedding networks cluster faces by demographic.

Minority groups experience reduced obfuscation utility.

Analytical model explains clustering effects.

Abstract

The proliferation of automated face recognition in the commercial and government sectors has caused significant privacy concerns for individuals. One approach to address these privacy concerns is to employ evasion attacks against the metric embedding networks powering face recognition systems: Face obfuscation systems generate imperceptibly perturbed images that cause face recognition systems to misidentify the user. Perturbed faces are generated on metric embedding networks, which are known to be unfair in the context of face recognition. A question of demographic fairness naturally follows: are there demographic disparities in face obfuscation system performance? We answer this question with an analytical and empirical exploration of recent face obfuscation systems. Metric embedding networks are found to be demographically aware: face embeddings are clustered by demographic. We show how this clustering behavior leads to reduced face obfuscation utility for faces in minority groups. An intuitive analytical model yields insight into these phenomena.