Communications of the ACM
Simple Make-Up Tips Can Help You Avoid Facial Recognition Software
Facial recognition algorithms can be foiled by a dab of strategically applied makeup that is subtle enough not to draw human attention.
Credit: Jochen Tack/Alamy
Researchers at Israel's Ben-Gurion University of the Negev developed artificial intelligence software that can provide makeup advice to foil facial recognition systems.
Tests showed the software's recommendations tricked real-world facial recognition systems 98.8% of the time, with the success rate in identifying women wearing its recommended makeup declining from 42.6% to 0.9%, and from 52.5% to 1.5% for men.
The adversarial machine learning system identifies which elements of a person's face are considered unique by facial recognition systems and highlights them on a digital heat map, which is used to determine where makeup can be applied to change the person's perceived face shape.
The system recommends only natural makeup hues, so people potentially could protect their privacy without drawing attention to themselves.
From New Scientist
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found