Sunday, July 03, 2005

ID Cards: Where Next?

Well, despite my repeated statements that we are going to win, the ID Cards Bill got through its second reading by 31 votes after some 40 MPs either abstained or didn't show up. Fix on that: they didn't bother to turn up. Gape at the depth of crapness. And write to them. Remind them that the scheme cannot possibly work. There are 44 million people who would need cards issuing. Assume that they must use the card once. The best performance achieved in the government's own ID test was a 96% success rate; or, to put it another way, a 4% failure rate. Engineers bandy around the phrase "five nines" to describe the level of reliability required in serious applications - that is, 99.999% reliability on 99.999% uptime. But even if that was achieved, even if the system was as reliable as the public-switched telephone network, there would still be 44 misidentifications in our very much watered down experiment. I've said this before, but it bears repeating.

Now consider that those 44 million people use the cards more often; Charles Clarke infamously suggested they might be checked when renting a video. That is silly, but it's no less silly than his suggestion that the best way to counter the emergence of a database society is to build a huge database of everyone in society. Say 100 times a year. 4,400 misidentifications. There are three possible endgames: the project fails before launch, amid total public humiliation and vast waste of money; it launches, misidentifications are frequent and people get locked up for no reason; it launches, misidentifications are frequent, and people simply don't check the cards, rendering it completely pointless. Of course, no.2 might lead speedily to no.3, and there's also the possibility that some other workaround appears, or widespread forgery, or the core database servers fail and take out the whole shebang. What our little exploratory calculation (serious engineers call it a "feasibility study") tell s us is that biometrics don't scale. This even goes for fingerprinting; get a big enough population and there will eventually be repeats, or at least cases similar enough to not be distinguished. The reason fingerprinting works, of course, is that as a rule you are only dealing with a tiny subset of the population, those who might reasonably have been at the scene of the crime at the time it was committed.

Biometric national ID's problem is that it's like investigating a crime solely by taking dabs and going house-to-house around the whole country looking for matches. It could either be effective and intolerably draconian, or ineffective and intolerably annoying. And annoying ineffectiveness is considerably easier to achieve.

No comments:

kostenloser Counter