Scientists fool a facial-recognition system with natural makeup

A group of scientists in collaboration with the Japanese NEC conducted a laboratory experiment, during which they attempted to fool the facial-recognition system by applying natural makeup to the subjects.

According to the researchers, natural makeup is hard to spot with the naked eye and its use would not raise suspicions.

“Our method detects natural makeup that, when applied to the attacker\’s face, hides his or her identity from the facial-recognition system,” they say.

The scientists loaded into the surrogate model photographs of «blacklisted» individuals to create a heat map showing the most important regions for identifying a given person.

Then, on these areas they applied makeup to alter the appearance of the participants.

For example, powder narrowed the nose, and contouring made the cheekbones more pronounced.

The researchers repeated this process until they managed to fool the model.

\n

In a small experiment conducted on ten men and ten women aged 20 to 28, facial-recognition cameras correctly identified the “blacklisted” people with makeup in 1.22% of cases.

The baseline model itself was not highly accurate. According to the study, it correctly identified participants without makeup in only 47.57% of cases, while with makeup applied, the rate dropped to 33.73%.

The researchers say that the drop in facial-recognition performance with makeup is below the realistic threshold for a real-world operating environment.

Earlier this year, scientists generated an image of a “universal face”, capable of deceiving most biometric identification systems.

In July, researchers reported the discovery of a method for unobtrusive insertion of malware into neural networks.

In May, researchers from the Maryland Center for Cybersecurity identified a vulnerability in neural networks that increases their energy consumption.

Subscribe to ForkLog’s AI news on Telegram: ForkLog AI — all the news from the world of AI!