Although the original Times article sensationalises the story by suggesting QinetiQ's millimetre wave imagers could be deployed across the whole of London underground at a cost of, ahem, £150,000 to £2 million per station, at a total of 270 stations, the core of the story appears to be confirmed by the denial, which concedes that QinetiQ has supplied unspecified "equipment" to TfL. Obviously, if TfL is looking at millimetre wave technology, it must be considering where it could be applied. But if it has been looking, it should have fairly swiftly concluded that widespread deployment would require radical surgery to its own operations, substantially increased staffing by both TfL and Transport Police, and huge quantities of money beyond the basic equipment costs.

Nor, indeed, does there seem any likelihood that such a system would be effective. Millimetre wave is currently used by UK immigration authorities as a people detector in order to track illegal immigrants concealed in lorries. Last year the Metropolitan Police also announced that it would be getting one, so it it doesn't have the kit already it will probably have it soon. The Met's scanner will we expect be as useless for the organisation as its existing low-dose x-ray scanner, which a while back we concluded had no clear purpose beyond psyops, and which the Met hopefully keeps offering to schools as a weapons detector. Millimetre wave might however be more useful at airports than x-ray, which is currently being tested at Heathrow, and the Times report points to a couple of other areas QinetiQ probably has in mind.

We think we can detect two distinct QinetiQ sales pitches, one aimed at mass transportation networks, and one that appears to be aimed at the Israeli market. The mass transportation one is by far the more costly and ambitious. Here, the scanner first picks up anyone carrying a suspect device, and sends off an alert to "nearby officers". The system can be linked to CCTV systems which "automatically pick out and follow the suspect until he can be stopped and questioned."

Such a scenario is conceivable, but the costs and difficulties swiftly become apparent when you considers how it could be applied to a mass transportation system of size and complexity. The costings given for London underground probably relate to the varying numbers of entrances (i.e. choke points that can be used for surveillance) at stations. Outlying stations may have only one of these, but further in there are several, and in central London there can seven or eight. Width of entrance and volume of traffic will also be issues, because whatever it is that is deployed to pick up suspects is going to have to be able to cope with hundreds of people a minute, many of them carrying packages, rucksacks, bags or briefcases.

The point here is that although it might be awfully clever to have a 'real time' (as QinetiQ claims) scanner than can see through clothes right down to the person underneath, the really clever bit is figuring out where the worrying stuff within the naked, rushing mass is. Shoving the 'modesty sticker' onto the right bits of all of the people (as QinetiQ security division MD Simon Stringer suggests can be done) would be pretty clever in itself, but getting the bombs, knives, pistols or whatever is a major challenge. Part of the difficulty lies in the application of the technology to what would effectively be a perimeter defence system, whereas existing (and more plausible) deployments tend to use it at established security choke points where it's simpler to zero in on individuals or areas.

QinetiQ seems to envisage using machine intelligence to do the zeroing in, and its example, the automatic detection of "the waistcoat bombs usually worn by suicide bombers" suggests the company might be seeking sales in Israel. Waistcoats are certainly frequently used by suicide bombers in Israel, but there are particular reasons why suicide bombing (as opposed to the non-suicide variety) is employed in Israel, and also why waistcoat bombs are used. Suicide bombing is less likely to be used in areas where the terrorist organisation does not have a steady supply of unskilled volunteers, and bombers will adopt the forms of concealment most likely to blend in. So, at an Israeli checkpoint a millimetre wave scanner preprogrammed to spot likely suspects might be useful if it reduces queuing times and thus reduces the likelihood of the bomber detonating in the queue, and if it also has a high probability of detecting a commonly-used method of concealment. There we also have the advantages that there is a queue of people expecting to be checked, and because the checkpoint is manned, the "nearby officers" really are nearby.

It is possible to program detection capabilities for likely objects and shapes into detector systems, but usually this process is helped greatly because detectors tend to be deployed in areas where the range of objects typically carried will be narrow, and where many of the concealment options are predictable. So at an airport relatively few people will be carrying things that look like guns or plastic explosives, while anybody trying to get guns or plastic explosives through will have a fairly limited number of options in terms of concealment. And everybody you're checking is going to be in an orderly line.

None of this is the case on London's underground, or on mass transit systems in general. In central London the system would have to check large numbers of people per frame, per second and the range of items carried is so varied that it would be beyond current technology, however much of that you threw at it. QinetiQ's claim that the system is "real-time" is also interesting in this light. It clearly can't be instantaneous, and although it may be faster than x-ray, it seems probable that detection speed will be related to the hardware and software deployed along with the scanner. We would expect that at say, Oxford Circus, Bond Street or Euston this would have to be amazingly heavy-duty stuff.

Then there's the question of that hand-off to CCTV, and then to security, which is floated as being possible (salesforce-speak for 'costs extra'). Aside from the complexity of framing and holding an individual for long enough to pass them across to the other system, you have the cost and complexity of that other system to consider. London already has a very high CCTV density, but these are usually used (despite what it often says on the tin about crime prevention) for forensic purposes; the CCTV footage is scanned after an incident, and the movements of suspects are, maybe, pieced together from that. Having machine-monitored CCTV track individuals is something numerous security services dream of, but it's not something that can be done practically today, and current CCTV control centres would need staff numbers and skills to be dramatically and expensively upgraded for them to stand a chance of 'following' an individual through a busy transport network.

And the nearby officer? For most of the London network, much of the time, there's scarcely anybody nearby apart from you and the other suspects. Outlying stations can have as few as one part-time member of staff, and even some quite busy ones fall back to a de facto honesty system outside of peak. Local police stations will have some ability to respond to alerts, but it won't be rapid response outside of central London, and even the concentration of London Transport Police in the central area would have trouble closing in on individuals before the tracking systems lost them. So what do you do then? Issue an alert and clear the station? Close the line? Close the network? But we shouldn't be too hard on London here - transportation systems throughout the world would face similar difficulties in trying to implement what is in effect a perimeter defence system on networks whose cost model depends on very low staffing levels and a high degree of honesty and self-policing on the part of passengers.

That's not to say they don't dream of an amazing technological fix to their current insurmountable problem at some point in the future. Also under consideration at international level, we are told, is a CCTV-based system that uses movement patterns to identify suspicious characters. So, if we take the example of standing on the platform waiting for the train, we can consider it possible that a potential bomber (or mugger or pickpocket) might follow different movement patterns to those of ordinary passengers. Obviously this is another one that will need a lot of clever IT underpinning it, and it might not turn out to be doable on public transport.

There has however been some research that indicates systems of this sort may have a value in narrower areas such as multi-storey car parks. Here, people who actually own a car in the car park are generally either walking towards it or wondering where the hell they left it, and a thief surveying the cars for likely victims would be unlikely to be able to mount a convincing impersonation of somebody wondering where they'd left their car. It's actually conceivable that an automated system here might be better at spotting thieves than humans would be. But mass transit implementations? Surely not for a good while yet.