News

net.wars: Biometrics for babies

This week's entertainment has been Biometrics 2005, a time when experts in this growing, formerly obscure field, exchange research notes. And basically you get the sense that all the stuff would work perfectly if you could just get rid of the human factor.

"People," one of the leading biometricians, James Wayman, said to me once,"just never have what you think they're going to have where you think they're going to have it."

It turns out that this applies especially to children.

Now, you might well ask whether we've fallen into a 1950s horror movie, but the EU has a resolution that requires biometrics on all travel documents issued by the end of 2006. The days when even small children could travel on their parents' passports is, of course, long gone, so this means that any travelling child will have to have its own passport. And, therefore, someone will have to capture its biometrics. How hard will that be?

Hard.

Everybody's trying to figure out how to meet legal requirements laid down by the ICAO for travel documents, but the question of whether capturing biometrics from children posed any special problems had been largely overlooked.

Enrolling – the technical name for the process of sitting you down and collecting your fingerprints, facial scan, or iriscode – is difficult enough with adults. Sweden, for example, is about to introduce live enrollment for its e-passport program at about 100 police stations, and the presentation about it was a litany of challenges: controlling the environment, training non-photographers take photographs, figuring out how to make people comfortable. They worked hard at it, and I guess we'll find out soon how successful they were.

Noticing that there was a lack of research on taking biometrics from children, Ruud van Munster, a senior project manager at the Netherlands' TNO Science and Industry , decided to investigate. He and his colleague Stephanie Moro-Ellenberger, recruited 161 children under 12 from colleagues, a day care Center, and an after school program – a small sample, to be sure – and experimented with taking fingerprints and facial scans. The results were startling enough that van Munster says he thinks the EU may reconsider and incorporate an age limit.

Of course, it was already known that humidity is a big problem in capturing fingerprints. Turns out this is worse with children: their hands tend to be both sticky and damp (you're a parent, you knew this). Thumb-sucking tends to have a negative effect (cripes, when I was small, it was only bad for your teeth; now it can make you a biometric outcast). However, the thumb is the typically the first finger that can be successfully enrolled, partly because it's bigger. Van Munster's recommended optimal strategy for conditioning fingers: wipe first with a wet tissue (to get rid of the stickiness), then with a dry Kleenex (to get rid of the dampness). The conclusion: fingerprinting isn't really possible for children under six. Quite apart from the problems listed above, there is the possibility of severe damage to the children due to extreme frustration on the part of the enroller.

"You really have to like two to three year olds," he observes.

Facial scans are a bigger problem. For one thing, it's hard to get children to focus on the camera and, when they do, to keep them focused on it long enough for the shutter to click; these systems use older digital cameras, which can take as long as a second and a half to react when you push the button. Moro-Ellenberger says that every time they got her one-year-old lined up in front of the wall ready for his photograph, he'd suddenly make a run toward the camera. Children need to be closer to the camera than adults because their faces are smaller; but putting them closer meant the focus was poor.

For another thing, children's faces change a great deal as they grow up, and quickly: a child's face changes substantially during the typical five-year validity of a passport. Girls' facial bone structure matures at about 14, boys' around 16. Worse, despite the existence of systems to predict what a long-time missing child might look like now, it isn't really predictable just from a single photograph how they will age: different parts of the face grow at different rates (the same is true of ears, by the way). One question the team tried to settle: babies all look the same, right? So their facial scans should become more individually distinguishable as they grow older, right? Actually, no: they look equally different to the facial scan system.

So, ultimately, does it make sense to require biometrics from young children? Not according to van Munster's research: he says fingerprints don't work for children under six. Facial scan was much more successful than fingerprints and at a much younger age. But under two years old even that was only 77 percent successful. That still compared favourably to getting fingerprints from the 48 children in the sample who were under two, which had a success rate of: zero.

What else did van Munster learn while trying to fingerprint children?

"Babies under nine months can make a very strong fist."