Privacy Visor

Privacy Visor Beats Facial Recognition Systems

posted by Paul Fiddian | 23.01.2013

Researchers in Japan have developed glasses that can defeat facial recognition technologies by emitting near-infrared light beams, invisible to the naked eye. Presented with this light source, facial recognition software can't accurately identity these glasses' wearers, it's claimed.

The so-termed ‘Privacy Visor' design is the brainchild of professors Isao Echizen and Seiichi Gohshi, from Tokyo's National Institute of Informatics and Kogakuin University, respectively.

Their aim was to counter what they describe as "an invasion of privacy, caused by photographs taken in secret." The pair's initial work in this field occurred after Echizen noted that Google Picasa's facial recognition system was effective even when subjects wore sunglasses and had their heads deliberately tilted. The scenario inspired them to explore how to create noise in between cameras and faces and they found their answer in near-infrared light, as produced by the Privacy Visor.

Facial Recognition Visor

Privacy Visor

‘Because this noise appended to the facial image causes a considerable change in the amount of features that is referenced at facial detection, facial detection is misjudged and recognition of people's faces is prevented', the Privacy Visor's developers explain in a press release.

Battery-powered, the Privacy Visor is made from plastic, with polycarbonate used in the lenses. The total LED count is 11, with eight placed around the eyes and a further three arranged around the nose.

With the Privacy Visor unveiled, Echizen and Gohshi are now refining the technology to produce a follow-on version that doesn't need batteries. They believe, once commercialised, these upgraded glasses could likely be produced at a per-unit cost of just $1. "We are developing an improved version of the privacy visor without power supply consisting of transparent materials that reflect or absorb specific wavelengths", Echizen explained in a statement.

In a report not connected to this research, ‘hacktivist' group Anonymous describes other ways of rending facial recognition software ineffective. They include applying heavy makeup and tilting the head back to an angle of 15 degrees. In both instances, the software won't register the presence of a face at all, says Anonymous.

Privacy Visor image courtesy Isao Echizen


Be the first one to add a comment

Add Comment

Fields marked with are required

Oops! Your comment was not submitted.

An internal error occurred during your comment submission. Please try again later or contact the system administrator.

The views and opinions expressed here are that of the individuals and are in no way related to Copybook Ltd.

Related Content

page up