This Facial Recognition Tech Can Not Be Deceived

This post is also available in: עברית (Hebrew)

UK government is interested in facial recognition technology in order to manage border crossings between Northern Ireland and the Republic of Ireland following the U.K.’s departure from the European Union. One of the companies that raised the government interest is Iproov, which has also won a contract from U.S. Department of Homeland Security to build a system using its facial recognition technology to bolster security and reduce waiting times at border crossings.

The company’s technology, which is already used by banks, can be used with a mobile phone and with pre-existing photographic records, such as passport or driving license photographs. It works by using the phone or other devices to take a short video of someone’s face while shining a pattern of colored light at it. It then analyzes reflection, comparing it to what its own system expects, to verify the person’s identity, as bloomberg.com reports.

Unlike other facial recognition systems, Iproov’s system cannot be fooled by someone holding a digital image up to the device camera. This is becoming a major concern as machine learning networks become ever-better at generating fake images.

It is also unique in being able to use photographic images from an existing database with the video images taken by the mobile device, without requiring a user to enroll their face using that device.

Iproov’s pilot with the U.S. border agency is part of a four-phase contract worth up to $800,000. The idea would be that travelers could use Iproov’s technology to “self-serve” the document check that normally happens at the border itself, authenticating themselves against a pre-registered photograph.

According to the company, their system had been shown in benchmark testing to be 100 times more accurate in checking someone’s identity against a passport photograph than a trained human passport officer.