LowKey: Leveraging Adversarial Attacks to Protect Social Media Users from Facial Recognition

TL;DR

This paper introduces LowKey, an adversarial filter designed to protect social media users from facial recognition by effectively disrupting industrial-grade systems and commercial APIs like Amazon Rekognition and Microsoft Azure Face Recognition.

Contribution

We develop a novel adversarial filter that accounts for the entire image processing pipeline and demonstrate its effectiveness against real-world facial recognition APIs.

Findings

Reduces Amazon Rekognition accuracy to below 1%

Degrades Microsoft Azure Face Recognition accuracy to below 1%

Provides an easy-to-use web tool for privacy protection

Abstract

Facial recognition systems are increasingly deployed by private corporations, government agencies, and contractors for consumer services and mass surveillance programs alike. These systems are typically built by scraping social media profiles for user images. Adversarial perturbations have been proposed for bypassing facial recognition systems. However, existing methods fail on full-scale systems and commercial APIs. We develop our own adversarial filter that accounts for the entire image processing pipeline and is demonstrably effective against industrial-grade pipelines that include face detection and large scale databases. Additionally, we release an easy-to-use webtool that significantly degrades the accuracy of Amazon Rekognition and the Microsoft Azure Face Recognition API, reducing the accuracy of each to below 1%.