How Foolproof is Almost Foolproof
Always reject answers that don’t have an estimate of error. That was drilled into me in the first year of High School Physics. And supported by many a mock exam (I missed the real exam, “The Wrath of Khan” was on the telly). So when I saw this article in The Register, and the Head of Metropolitian Police quoted as follows, I have to ask what the estimate of error is.
“ID cards can only be the answer if the recognition of them is almost perfect,” Sir Ian said, adding that the technology had to be “as close to foolproof as possible”.
Okay, so simple question. What’s the difference between perfect and almost perfect? Is almost perfect 99.99%? 99.9%? 99%? 9%? Or to put it in trying not to hide how bad it is, 99.5% (which sounds really really great) means an error rate of one in two hundred. That’s a great reliability rate. Let’s imagine a nightclub of four hundred people, and the police thinking that maybe somebody “evil” is inside. Which in the current climate could eaisly happen in certain areas of large cities. Two people from the nightclub will be “falsely read” and probably have to spend a night down the station sorting out who they are.
Of course, that’s assuming the database is perfect… and that your bloodshot eyes aren’t screwing up the scanner… or the spilled nail lacquer on your index finger gets past the fingerprint scanner… and that there’s no problem with the GPRS data link back to the single central server that checks for a forged card. Oh yes and the Government says you’ll pay £93 for this ‘privellege.’
Oh it must be very accurate then, you’re thinking. Here’s numbers from one official trial.
Iris Scan… 96% success (1 in 25).
Fingerprint scan… 81% success rate (roughly 1 in 5).
Facial scan… 69% success rate (roughly 1 in 3).
Fancy visiting No2ID yet?
we’re coming for you citizen, please stop thinking and wait for our call.
its very very worrying isnt it ;-(
Funny people, you are… ;)
We’ve had ID cards for almost forever, with no issues that I can think of. Get over it.
The *really* fun bit is that the errors could add up, so that out of our 100 people, 4 would be misidentified by the iris scan, 19 by the fingerprint scan, and 31 by the facial scan.
I make that a 54 people, i.e. a 54% failure rate, and that’s less reliable than tossing a coin. Not very useful in say finding a “bad man” in a football crowd of 80,000 people…
Jim, the great thing is if I go along to be mass-processed at one of five buildings in the UK, then teh scanning rate at teh centre is surely going to have a failure rate as well. How much is that? Surely not the same as the on the street readers?
And law or no law, someone wants to shine a laser in my eye to read the blood vessels? Riiiight… I can’t watch someone get eyedrops in, let alone do it to myself.
What is that officer? Don’t stare into the class 4 laser light with my remaining good eye.