This Facial Recognition System Can Identify You Even If Your Face Isn’t Visible

A new facial recognition system can identify who you are without needing to see your entire face.

Created by researchers in Saarbrücken, Germany, the Faceless Recognition System recognises similarities and patterns around people’s heads and bodies to identify who is in the photo, even if their face isn’t visible.

Speaking about the system, the researchers from the Max Planck Institute for Informatics in Germany said: “From a privacy perspective, the results presented here should raise concern.

“We show that, when using state of the art techniques, blurring a head has limited effect.

“We also show that only a handful of tagged heads are enough to enable recognition, even across different events (different day, clothes, poses, point of view).”

How well the system works depends on how many visible, unobscured faces there are in a data set – the study finding that if someone’s face was shown 1.25 times in a set, the system had a 69.9% accuracy rate in identifying them.

If their face was visible 10 times, however, the Faceless Recognition System had a 91.5% success rate in identifying the person.

The authors of the report said: “In the most aggressive scenario considered (all user heads blacked-out, tagged images from a different event), the recognition accuracy of our system is 12× higher than chance level.”

That said, if someone’s face was blocked out by a black square in multiple photos, the system’s accuracy dropped to 14.7% from 47.4% – a figure that is still three times higher than systems using blind prediction.

For the research, the team used 40,000 Flickr images of 2,000 different people, testing the system when their faces were both fully visible and completely blocked out by a black square.

And the team said that this type of algorithm is likely already being used online, saying: “It is very probable that undisclosed systems similar to the ones described here already operate online.

“We believe it is the responsibility of the computer vision community to quantify, and disseminate the privacy implications of the images users share online. This work is a first step in this direction.”

To read the full report by Seong Joon Oh, Rodrigo Benenson, Mario Fritz, Bernt Schiele please click here.