UNC researchers suggest that certain facial recognition technology can be easily traversed

UNC researchers suggest that certain facial recognition technology can be easily

Researchers in the security and computer vision fields from the University of North Carolina (UNC) have recently revealed the drawbacks of the use of biometrics – like facial recognition technology – for the purpose of digital security.

According to the details shared by the UNC researchers at the Usenix security conference held earlier this month, it is fairly easy to traverse certain facial recognition technology with the help of some 3D rendering and Facebook stalking.

The researchers further explained that, on the basis of the results of their research, they have found that most of the existing facial recognition systems can be tricked by the virtual reality (VR)-like computer-rendered faces which they have created.

For its research, the UNC researchers’ team collected photos of 20 volunteers from their Facebook page, LinkedIn profile, and other online sources. Then they created 3D renderings of the faces of the volunteers, added facial animations, and made slight adjustments to their eyes to look straight at the camera.

As a result of their efforts, the researchers were able to successfully -- 55-85 percent of the time -- fool four of the five of facial recognition systems which they tried to hack. The five systems tested by the researchers were Mobius, TrueKey, KeyLemon, BioID, and 1D, which are available on both the Google Play Store and the iTunes Store.

economie: 
emplacements: 
Share Share