There have been plenty of indications of how facial biometric technology has advanced in recent years, but one of the more unusual new ones is a system that can identify pixelated faces from digital images.
The system was developed by a team of researchers from the University of Texas, and tested against major privacy protection systems including YouTube’s face-blurring function. It’s based on standard machine learning tools that are widely available, according to the researchers, whose aim, in part, is to raise concerns about online privacy protection.
There are some notable shortcomings of the system, however. For one, the system needs to be trained to look for the faces of specific individuals. And while its success rate in identification could reach 90 percent in some applications, it also was as low as 17 percent elsewhere, as when attempting to identify celebrities faces obscured by the P3 encryption mechanism used for JPEG photos.
Still, it’s a useful reminder of the rapid advance of facial recognition technology, and to privacy advocates, a warning about further battles against the technology’s encroachments as its deployments become more widespread.
September 15, 2016 – by Alex Perala