Home / News / Technology news

Techno News


Scientists deceived the face recognition technology

A group of scientists from the Max Planck Institute for Informatics (Germany) has published an article on a scientific resource arXiv, where they describe the way to deceive face recognition technology. However, they were not trying to create an image that would be recognized as a real face, but quite the opposite. According to the researchers, many social networks’ users are concerned that they could be recognized on images, while they would prefer to stay anonymous. The simplest methods are used for protection in these cases such as placing black squares on the face or blurring facial features. However, the first technique makes pictures look weird, to put it mildly; while the automatic facial recognition technologies easily overcome the second one.

Scientists approached the issue quite seriously and developed a technology to create “fake faces”. The captured faces are reproduced in a form of a point cloud of landmarks. After that, the sequence of these points is changed in a certain way. Next, the real and generated pairs of points are put into the “head generator” and the program creates a new face. According to the authors of the paper, this algorithm can be used on any images, no matter the head tilt, lighting conditions, background, etc. While the result is an image looking at which users of social networks won’t see what’s wrong at first. However, the modern face recognition systems won’t be able to tell the user.

What’s unclear though is the practical value of this work. It’s obvious that if a person cares about his or her confidentiality they simply won’t upload pictures of social networks, on which they don’t want to be recognized.

Back to the list

Scientists deceived the face recognition technology