26 September 2023

Losing face: How facial recognition tech can’t face up to crowds

Start the conversation

Avery Thompson* says facial recognition is increasingly being used to search for criminals in crowds, but it has an awful problem with false-positive identifications.


Photo: Rishabh Varshney

Facial recognition technology is increasingly being used to identify people in public places.

It’s used in China to identify fugitives and jaywalkers, in Chicago to spot wanted criminals, at the Olympic Games to track journalists and event officials, and even inside Amazon stores to track purchases.

This raises a whole bunch of complicated privacy questions, but that’s not all.

Data from facial recognition systems in the UK shows the danger of false-positives.

Facial recognition systems operate on a very simple premise: cameras can scan crowds of thousands of people continuously, and software can pick up on certain facial characteristics to positively identify an individual.

For situations like stadiums or crowded public squares, facial recognition is far more effective — at least theoretically — than patrolling an entire crowd with human police officers.

The South Wales Police recently implemented a facial recognition system at the UEFA Champions League Final, where the system identified 2,470 suspected criminals.

That sounds pretty impressive, except that all but 173 of those suspected criminals turned out to be innocent people.

That’s a 92 per cent failure rate.

The South Wales Police blamed this statistic on bad camera equipment and a first-time implementation, but in subsequent outings the system hasn’t got much better.

Over the 15 events in which the South Wales Police have used facial recognition, only five saw more correct versus incorrect identifications.

Only three events had zero incorrect identifications.

Plenty of people have already criticised facial recognition on its privacy aspects, and now there’s also the possibility that you could get interrogated by police because a computer mistook you for someone else.

Fortunately — or perhaps, unfortunately — the technology will likely improve over the next few years and mistakes will become more rare.

But that’s small consolation for the thousands of people falsely identified by the South Wales Police alone, and the possibly tens of thousands experiencing that same fate around the world.

Perhaps someday facial recognition won’t make this kind of mistake, for what little comfort that’s worth.

* Avery Thompson is science, energy and tech reporter for Popular Mechanics. He tweets at @PhysicallyAvery and his website is averythompson.net.

This article first appeared at www.popularmechanics.com.

Start the conversation

Be among the first to get all the Public Sector and Defence news and views that matter.

Subscribe now and receive the latest news, delivered free to your inbox.

By submitting your email address you are agreeing to Region Group's terms and conditions and privacy policy.