Going slow, and going underground
November 16th, 2005Sorry for not posting of late, but it is for a good reason - I have a new job. And it’s going fine, although I am having to remember many long-forgotten Unix commands… Anyway, while I’m still finding my feet, I won’t be blogging much.
But here’s a quickie. Earlier in the week Alistair Darling discussed introducing additional security measures on trains. So, we’ll be scanned, have sniffer dogs set upon us, have our bags searched, just like on the aeroplanes (although it will be random rather than compulsory). We’re even getting ‘intelligent’ CCTV cameras, that can “detect” faces*. Except of course, that unlike aeroplanes, no-one’s ever steered a train into a skyscraper and blown it up. Although the high mass and speed of a train make it an initially attractive source of kinetic energy with which to wreak destruction, trains are not really ideal to hijack, what with the driver never actually being in full control. I’d be fairly confident in saying that a train is pretty low down the list of terrorists’ preferred options** for taking over for nefarious purposes.
Of course, trains can still be blown up and cause considerable carnage, as Madrid and London have seen, regrettably. But, this is still no good reason for the latest proposals. This is not because of the impracticalities - they would have to scan three million people a day on the Tube alone; instead it’s the sheer frigging dumbness of the ’solution’. Trains may be labelled a “soft” target by Darling, but so are cafés, restaurants, shopping centres, shops, pubs, offices, parks, pedestrian crossings… in fact, practically anything in this country. Posting millimetre-wave scanners on every ticket gate isn’t going to stop the truly deluded and hateful from commiting murder; they’ll just look elsewhere (Israel, with its fortress of a public transport system, is a case in point - the suicide bombers just went elsewhere). Perhaps instead we should start thinking about terrorism as something more than what can merely be solved by throwing technology at it.
* Incidentally, a BBC report quoted that: “a few years ago, it took 10 or 15 seconds to compare a face. Now it takes three”. That’s not at all encouraging - it implied that the systems have Moore’s law to thank, rather than any actual advances in the field.
** Unless the only Western film they have ever seen is The Taking of Pelham One Two Three
November 16th, 2005 at 23:38:48
On face recognition: it takes 3 seconds to check someone’s face against a single face record, so if you have 10 suspects then that is 30 seconds per person. Then when the software recognises someone (probably incorrectly) there will be a mad scramble to find them amongst all the people on the underground who have gone past that ticket barrier in the last minute or so. What is wrong with the old-fasioned approach of putting suspects under surveillance?
Max
November 17th, 2005 at 09:03:56
Intriguingly, the actual quote says ‘”A few years ago, it took 10 or 15 seconds to compare a face. Now it takes three,” he said. “But if you are comparing that face with, say, 20 terrorists’ faces, that’s 100 seconds, so it becomes impossible.”‘ Don’t know about you, but if they’ve got this sort of problem with basic arithmetic, what hope for the application of technology to the issue?
November 17th, 2005 at 15:37:18
Meanwhile they will carry on trundling nuclear waste across the capital on the north london line, presumably. Because there’s nothing dangerous about that…
red
November 18th, 2005 at 14:34:15
Let’s not forget that the middle of a big crowd of people being scanned for bombs is an ideal location for any prospective human bomb to set him/herself off. You don’t need to be on the train to cause immense damage. In fact, the stairs going into the station would do. Bring down the whole of the entrance area, damage the stairs, lift, escalators whatever, and trap everyone in the station.
December 6th, 2005 at 19:17:42
They’re testing ’sophisticated CCTV’, no mention of us getting it as you suggest. This testing has been going on for a long time without positive results. Crowded environments and simple processing power limitations are partially to blame alongside the impossibility of the project’s requirements (set that high just to get funding). 3 seconds for one face does not translate to 30 seconds for 10, even the most moronic developer will use something better than a linear search, and the time for a single face is well below 1 second for most software. However, when the face is 20 pixels high in a crowd you’d be hard pressed to get a reliable recognition at all, let alone recognising every face in the scene for every frame of video in real time. It’s madness I tell you, and a money pit.