The robot at the border: UK bets big on face scanning

Home Secretary Jacqui Smith is taking risks with public safety, whilst simultaneously condemning thousands of airline passengers to long delays this winter. That is the fear expressed by a spokesperson for the Public and Commercial Services Union (PCS), in reaction to government plans to test new face recognition software at airports.

From this month, the UK Border Agency is trialling new technology at Manchester Airport (and Stansted, according to the PCS) which is claimed to balance high security with quicker times at immigration control. New facial recognition gates will use scanning equipment to compare the faces of UK and EEA passengers to their biometric passports. If successful these gates could be rolled out across the country.

As we understand it, the approach applies only to those individuals who hold e-chipped passports or, we estimate, around ten to 15 per cent of the travelling population. On arriving in the UK, instead of passing their passport to an Immigration Official for the usual baleful once-over, new arrivals will place their passport on a scanner.

This will check the ID of the passport holder against a list of passengers flagged up by e-Borders as barred from flying into the UK. It will then access photographic data held on the chip, so that when the passenger steps forward to have their face scanned, a second “visual” check may be carried out.

The entire operation will continue to be supervised by e-Borders staff, who will now glare at passengers from a slightly greater distance. They may intervene at any time, and will pull a random quota of individuals from the queue for further checking. As far as the Home Office is concerned, this is an additional layer of security – not a replacement for what is already there.

There are several serious obstacles to the smooth running of this plan.

First are concerns by the Biometrics Assurance Group (pdf), reported previously in El Reg, that there is still work to do on both the facial recognition standards and the format in which facial images are stored.

It is not clear that this issue has been resolved – although the Reg has been informed that the software recently had to be recalibrated, because it was rejecting too many individuals.

This brings us to the second problem: all recognition systems throw up a number of false positives and false negatives. The precise number of each will be determined by an operational decision on where to set the acceptance criterion for individuals. A “risk averse” policy will increase the number of false positives (people stopped for no good reason), while a “risk-tolerant” policy will increase the false negatives (dodgy individuals allowed into the country).

This is a difficult call to make. One argument advanced by the Home Office is that this system will speed up passenger through-put: a few seconds per passenger, but a very large improvement when multiplied by the millions that arrive every year. It would be a stroke of wondrous serendipity to discover that the point at which this software saved all this time was also, magically, the point at which it identified as many potential terrorists and criminals as the present human method - or more.

Although in fact, for it to do that, it would need to be detecting wrong-doers far more effectively than the present system does. According to the PCS, no such comparisons have been carried out – or at least none that the Home Office has yet shared.

Computer recognition systems can work, but most have a lot to do before they even get close to matching the results from a human operator and good old-fashioned experience and intuition.

Which brings us to difficulty number three. The trial will take place in a live setting. If the software works, no problem; if it doesn’t, then expect Manchester to become destination number one for criminals seeking to avoid detection on their way into the UK.

Again, the Home Office responds that this technology has been used in numerous locations around the world, with “no major problems” reported. However, the calibration issue is one that is itself very risky to resolve. Too many false positives, and the Home Office can expect letters from outraged citizens.

Even a couple of false negatives, and the next we will know about them is when a major incident goes down somewhere in Central London.

And finally - the PCS themselves are advising staff to have nothing to do with these trials. That is unlikely to have much of an impact. Not all Immigration Staff are members of the PCS (many belong to an unaffiliated in-house body called the Immigration Services Union): nor will the PCS go so far as to declare official action over this issue.

But it won’t help.

One more time, we worry, as David Davies opined back in July, that Nu Labour are addicted to untried hi-tech solutions, taking risks with public safety that a more measured, less headline-obsessed approach would avoid. ®