Computer scientists at Israel's Tel Aviv University (TAU) say they have developed a "master face" method for circumventing a large number of facial recognition systems, by applying artificial intelligence to generate a facial template.
The researchers say the technique exploits such systems' usage of broad sets of markers to identify specific people; producing facial templates that match many such markers essentially creates an omni-face that can bypass numerous safeguards.
The researchers created the master face by plugging an algorithm into a generative adversarial network that builds digital images of artificial human faces.
The TAU team said testing showed the template was able unlock over 20% of the identities in an open source database of 13,000 facial images operated by the University of Massachusetts.
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA