You don’t have to wear a Halloween mask to avoid being detected by facial recognition software: a dab of makeup will do, according to a new study.
Researchers at Ben-Gurion University in Beersheba, Israel, have developed artificial intelligence that shows users where to apply foundation or rouge to trick facial recognition algorithms into thinking they’re looking at a different person .
The researchers tested their scheme against ArcFace, a machine learning model that takes two facial images and determines the likelihood that it is the same person.
The team used 20 volunteers (10 men and 10 women) in a real environment using two cameras and a variety of lighting conditions and shooting angles.
Participants wearing “contradictory makeup” as recommended by the program cheated the system 98.8% of the time.
Normally ArcFace correctly identifies a face about half (47.6%) of the time, but this dropped to 1.2% with makeup.
The researchers called it “below a reasonable threshold of a realistic operating environment”.
Scroll down for video
Computer scientists at Ben-Gurion University have developed an AI that could recommend subtle makeup applications to trick the facial recognition system. The AI created a digital heat map of a participant’s face, highlighting areas that face recognition systems would likely be recognizable.
Facial recognition programs examine which parts of a person’s face are most identifiable.
Data scientist Nitzan Guetta and his team attempted to emulate this approach, then figure out how to subtly alter this unique characteristic without drawing too much attention.
AI developed a digital heat map of a participant’s face, highlighting areas that face recognition systems would likely be considered highly identifiable.
A digital makeup projection then showed where to apply makeup to alter the contours of the person’s face.
A participant wearing makeup recommended by the AI program (top) managed to trick a facial recognition system. His makeover had an intensity score of just 1.81, compared to 2.30 when he applied random makeup.
According to their study published on the open source Arxiv server, putting on makeup reduced the success rate of facial recognition from 42.6% to 0.9% for them, compared to 52.5% to 1.5% for men.
“Facial contouring may become more than a TikTok trend, and such makeup may become the next wave in protecting one’s privacy in public from automatic facial recognition systems,” said Mariann Hardey, professor at the School of commerce from Durham University. New Scientist.
While the participant is initially recognized by the FR system (top), when applying contradictory makeup (bottom), the system fails to spot him.
The intensity of the metamorphosis was also recorded, and the average intensity score of the contradictory makeup was lower than the random placement (2.64, vs. 2.85), “while being more effective at dodging the [facial recognition]”the researchers said.
The adversarial system has been programmed to stick to natural tones when formulating recommendations.
There are other makeup techniques, as CV Dazzle, which use intentionally jarring patterns, contrasting colors, and asymmetry to get rid of facial recognition technology.
But such extravagant makeup draws more attention to the wearer in the real world.