acm-header
Sign In

Communications of the ACM

ACM TechNews

Glasses Make Face Recognition Tech Think You're Milla Jovovich


View as: Print Mobile App Share:
The glasses fooled a computer into thinking a white male researcher was actress Milla Jovovich.

Eyeglasses with patterned frames developed at Carnegie Mellon University can obscure the identity of the wearer to facial-recognition systems.

Credit: Screen Gems

Carnegie Mellon University researchers have designed eyeglasses with patterned frames that can obscure the identity of the wearer to facial-recognition algorithms.

The team says the glasses were developed to mislead facial-recognition programs that use neural networks, which focus on pixel coloration and make comparisons with other, similar images to identify the subject. If a small area of the face has been altered, the system's ability to read the pixilation can be disrupted.

The glasses' patterns change how the system interprets the wearer's face by overlaying the face with additional pixels and causing the software to misidentify the person as another face in its database. For example, a white male researcher was able to pass for actress Milla Jovovich, and a south Asian woman successfully impersonated a Middle Eastern male, tricking the facial-recognition software Face++ 90% of the time.

As facial identification systems become more prevalent, the researchers say technology such as the patterned glasses could help protect an individual's privacy, but they also warn it also could lend itself to aiding criminals committing identity fraud.

From New Scientist
View Full Article

 

Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account