Facial recognition and security – is America ready?

The controversial technology has come of age - but have we?
19 December 2022

Facial recognition – a safer two-factor authenticator?

Getting your Trinity Audio player ready...

Facial recognition is a technology we think we understand, though most of what we know comes either from fiction or from its very early days – and potentially its premature deployment decades ago. Now, the technology is back in the news, having been used at stadia during the Qatar World Cup soccer tournament.

We sat down with Terry Schulenburg, VP of Business Development at CyberLink, the leading facial recognition company behind the FaceMe app, to ask whether it was time the technology was more widely deployed in the US.

THQ:

Does facial recognition work to the degree of confidence that we need it to for applications like security?

TS:

Define security.

I mean that very seriously, because most people approach facial recognition from the TV and movie perspective. I’m using it to capture bad guys, I’m using it to find the bad actor or I’m using it to do these other things. Security is much greater than that.

Let’s take airports, for example. Today at an airport, in the secure areas, anywhere that a flight attendant would go or a pilot would go, they use two-factor authentication. They use a card that they swipe, and then there’s a pin number that goes along with that. Now we all know how safe and secure that is, because nobody can steal the card, and nobody could ever learn your pin number.

Building a better door.

THQ:

Ahem…

TS:

As I say, we know how safe and secure that is. What happens when you add biometrics to the mix, and I’m specifically talking about facial recognition, which we’re doing in every airport in South America right now? Every airport has two-factor authentication at the secure areas. And they’ve been using card and pin for years. Well, now they’re using face and pin. And what happens now is if you don’t have the right face, you are not getting through that door. So the level of security now has gone up tenfold from having a card and a pin, where I can easily duplicate the card, or steal the card, and I can find a way to get your pin. So does it work to the degree of competence we want to use it for security? Absolutely.

I can take you to medical facilities that are using it to dispense drugs, I can take you to locations all across the country, nuclear facilities that are using it for two-factor authentication to get in the door. You don’t belong in a nuclear facility if you shouldn’t be there, and we’re only letting people in that belong there. So from the security standpoint, yes.

When you’re talking about stadia, this is a really interesting proposition. They’re doing facial recognition at the World Cup right now in Qatar. Are they able to catch bad actors?

Absolutely not.

That’s not what facial recognition is used for. And at the World Cup, it can’t do it, it’s so easy to sidestep facial recognition if you want to do it. It’s meant to keep people safe, it’s meant to keep your credentials safe.

“Open sesame.”

So when you want to use something and be a part of it, you can do it. Your banking application is a perfect example. When you use your banking application, you can type in your password. The banks have seen over the last couple of years that people want to use their face instead of their password, because it’s fast, it’s secure, and no one’s going to fake it out. That password can be stolen and used, and then thieves have access to all your money.

So when you talk about security, I have had to be really careful. Am I going to be able to find crooks in a crowd as they’re entering a stadium? No, that’s not what facial recognition is going to be used for. Could it do it? Possibly. In some instances, you might catch somebody in a wrong location, or in a bad position, and it’ll mistakenly say that they’re a bad actor or a criminal.

But the vast majority of the time, that’s not what it’s for. It’s meant to protect the athletes. The athletes are using it to get into the locker room — you belong here, you’re going get into the locker room. If you don’t, if you’re not an athlete, or you’re not a coach, you’re not getting in that locker room, no matter what type of credentials you have. And at the World Cup, that’s how the stadiums are protecting the people and the infrastructure around them.

THQ:

So facial recognition is an unbreakable, unstealable version of the definitive you, like multi-factor authentication that can’t be faked?

TS:

That’s the idea.

Thank you, Hollywood…

THQ:

Is America…ready…for that?

TS:

No.

THQ:

Oh.

TS:

One of the problems that the US has with facial recognition is that we have so much bad information about it. One of the things people have a problem with is “If you have my face on file in your database, can somebody steal my face and use it to open my bank account? “

The answer is no. But we see people do it all the time in the movies, we see them do it all the time on TV. And that’s the problem — there is so much misinformation. And people are just starting to learn, “Wait, that’s not the way that this works?”

Doing math on your face.

That’s not the way that the system is functioning. I’ve been working within the facial recognition space for almost 15 years. 15 years ago, we would have a face in the database, literally a picture of your face in our database. And when I captured a face on video, I would compare the two and I would have a confidence level that this is probably you. Now, the problem with that is you have lighting, you have sunlight, you have weather, you have fog, you have all of these things that can affect that confidence level. And all of those things affect that type of comparison.

Well, when you add artificial intelligence into the mix, things change. We’ve trained our system on over a million different faces. And what we’ve done is found the particulars of what makes a face different between you and me. And so what happens is I give you a score, I do math on your face, and I create something called a template. And that template is the only thing that’s stored in the database.

Now, and this is the crucial part, you cannot take that template off my database, and ever recreate your face. So no-one could use it anywhere else to unlock their system. It is something that only works in the environment that I have registered for. There’s an enrollment process that you go through. And that enrollment process creates that template. And then when I find you on a video screen, whether it’s in a security camera, whether it’s at a door trying to get access, when I grab that face, and I use a template to compare it, there are nuances. And those nuances are where AI comes in and takes the place of the human eyeball comparing two actual images. AI’s not comparing images with images, it’s comparing an image and a template, and it’s reading the nuances we wouldn’t normally pick up on just by looking.

Using AI, I can give you a ten year old or a 15 year old picture of myself. If I enroll that 15 year old picture, it will recognize me today based on those characteristics and nuances. And again, it’s not something that I can take or rehash. It’s a mathematical model, so I can’t steal it. I can’t recognize it as a human being. There’s nothing I can do with it. All I can do is compare it. And all I get is a confidence level back — thumbs up, thumbs down. Yep, we think this is Terry, for 98% sure. And that’s the level of confidence we’re dealing with.

I have to keep telling this to people because they’ve been misinformed by movies. Then people hear that and say “Oh, so you’re actually not storing my photograph anywhere?” And I’m not.

Facial turnstyling.

I just put this in at a large company. They’re using this at their turnstiles at their front door for their employees. They have 4000 or 5000 employees coming in every day. And they’ve had five turnstiles for years using cards. They put facial recognition on one of the turnstiles about a month ago. The first week, only about 10% of people used those facial recognition turnstiles. People saw them just flying right through, not having to pull out their card, not having to wait.

Two weeks later, those four turnstiles that were still activated by card weren’t being used. 90% of people were using the facial recognition. Why? Because it was so much easier. They didn’t have to remember their card, they didn’t have to find their card, they didn’t have to wonder where it was when they went to lunch, they didn’t have to remember to bring their card with them when they went to the bathroom. They were able to move about the building significantly more easily than they’ve ever been able to before. And that’s where we’re seeing facial recognition really start to take off in a secure environment.

 

In Part 2 of this article, we’ll explore the edge cases of facial recognition, and the continuing battle for hearts and minds over a technology that has suffered by being the perfect hook for espionage and dystopian fiction.