Wednesday, May 25, 2005

No More Biometric Trials

The Home Office has just released a statement concerning the reintroduction of the ID card bill.

The first line reads: "Please Note: There are currently no plans for further biometric trials."

Apparently the results of the biometric entrolment trials are vailable on the UK passport service website. I can't find them - but I'll keep looking. They may be published later today.

There is a brief summary of the trials in the Home Office statement - it does not mention the technical aspects of the trial, only people's perceptions of the process itself and their attitude towards biometrics - is this because that is the only good news for the government?


P.S. The spelling mistakes in the statement do not inspire confidence.


Charlie Williams said...

I've found the reports. They can be found under the "Publications" section of the UK Passport Service website. So, not obvious then.

There are a couple of things about this report. It states that it is not a technical trial. It was not the purpose of this trial to test the effectiveness of biometrics for any purpose - merely how the public reacted to them.

It states that further trials are needed to solve problems even within the limited remit of this trial, never mind the technical aspects.

What few technical aspects are available are truly shocking. For the purposes of ID verification iris scans were best with a success rate of 96% (for non-disabled), fingerprints second with 81% and facial last with 69%. Just imagine trying to prove your identity with success rates that low - and remember the verification process happened within minutes of the original ID registration.

Disabled people were even less likely to be correctly identified: iris 91%, fingerprint 80%, facial 48%.

As to the results of the survey - the sample was largely self selecting. Although MORI attempting to ensure a balanced sample by matching samples to the census the one question they did not ask was political allegiance - the one question which is most likely to reveal any bias toward or against ID cards.

In addition, of the 28 questions asked only four (one was a six part question) made any reference to collecting biometrics as part of an ID card. The remaining questions concerned ergonomics.

The fact that the report claims that there was 31 questions while the sample questionaire in the appendix only gives 28 is worrying too.

To the question "Where would ID verification facilities be appropriate?", none of the opportunistic group (7,266 people) chose nowhere - suggesting that they were all in favour of ID cards and represent a hopelessly skewed sample. The quota (chosen by MORI) group had 5% suggesting (from comparison with previous surveys)that they too were skewed in favour of ID cards. Further questions only helped to reinforce this perception when compared against more randomly sampled polls.


Watching Them, Watching Us said...

The slowness of the Verification times are astonishing !

Imagine being the last person off a Jumbo ket with 400 people ahead of you in the queue. Average Verification times of between 40 and 80 seconds are at least an order of magnitude too slow to prevent massive queues.

The worst cases took on average 10 minutes each to verify !

This is in tightly controlled Passport Control, where you can assume the ideal lighting conditions.

This will be much much worse in any otthe ID card Biometreic Reader at a bank or hospital etc.

Scribe said...

One also wonders what contingency plans are if someone fails their check. Assuming that the most reliable form of identity is used (iris scanning) and that there are no disabled people in the queue, if there are 400 people ahead of you, that means that 16 people of these will fail - what will happen to them all?

I suspect any real answer to this depends on the nature of the failure. Not matching your stored data is the most likely (false negative? I always get confused...), and could probably be solved by scanning again. A match against a "watched" identity could cause more problems...

Furthermore, I'm surprised 1 in 5 fingerprint checks fails - isn't the most mature use of biometrics? Either way, I'm not inspired just yet.

Charlie Williams said...

Fingerprints are an interesting case. They have been in use a long time for the purposes of crime detection but it has never been successfully proven that they are, in fact, as unique as is claimed\assumed. There was an article on this in New Scientist recently (but I can't find it by a quick search).

Some people do not have distinct fingerprints which makes it hard for biometric systems to operate effectively - for instance heavy manual labourers tend to have fingerprints that have been rubbed away so much that the tips of their fingers are essentially scar tissue destroying any trace of the underlying print.

Interestingly, one of the conclusions that was reached by the study was that a way needed to be found to take the prints of people with fewer than ten fingers - apparently the scanning system got confused with fewer fingers! One wonders what it would make of Hannibal Lecter, who had six fingers on each hand (yes, I know he's fiction - but the syndrome that caused it isn't).


P.S. Scribe, although iris scans are the most reliable to verify they are the hardest to capture with a a success rate of only 90% (for non-disabled people) - so of the 400 people in the queue 40 will be forced to use a less reliable system (fingerprint with 100% capture rate and 81% verification reliability). This means that 22 people out of 400 will fail a single biometric check. For the group of those unable to provide an iris scan 6% will not be able to prove their identity with the remaining two checks (31% failure for facial X 19% failure for fingerprint). Out of 49,000,00 people (pop of the UK over 16 to be issued with an ID card) that's 4,900,000 unable to provide an iris scan and 288,610 people unable to prove their identity - and that's ignoring the disabled population for whom the results are even worse.

It also ignores the idea that those likely to fail one biometric are also more likely to fail the rest of the biometrics as well - all of the tests require staying still for a period of time so those with motor related diseases or conditions (even if not classified as disabled) will fail to provide biometrics on all three tests.

There is also a racial element to biometric capture - some races (blacks) exaserbated by poor lighting condition can reduce the capture rate (for facial scans) to 22%, even though the capture rate for other races is somewhere around 60% (+- 10%) for the same poor lighting conditions. Iris scans are also less relaible for blacks.

Oh, and because it's a one-off study it doesn't explore how biometrics deteriorate with age and certain illnesses\diseases.

All of which backs up the various studies that have argued that biometrics will fall foul of the various legislation to prevent discrimination on the basis of race, colour or creed.

Watching Them, Watching Us said...

The first line reads: "Please Note: There are currently no plans for further biometric trials."

This Home Office Press Release now seems to have been censored with this line removed !

Charlie Williams said...

Now if I was cynical I'd say this could be part of a government plan - after all arguing against ID cards purely on the basis of technology is inherently flawed since technology improves all the time.

With the announcement of the new bill and trial results on the same day - the entire anti-ID community has concentrated on the trial results, since to them the bill is old news (unchanged from last time). But what the public at large sees is the anti-ID camp arguing purely on the basis of technology since they are unaware of the many critisisms of sections of the bill itself.

Now if the government can produce a better biometric trial* then our argument /appears/ to have been negated.

Any technological argument is, by definition, a red herring and could work against us in the long run forcing us to switch arguments halfway through.

We need to get back to discussing the problems with the bill, not the technology, and fast.


* hardly difficult given the results of the last one.

Scribe said...

I have to agree, although choosing a single argument at any one time is bound to fail - as we've seen, the government simply switch their pro-ID arguments, dismissing any criticism pointed at their last propaganda (is it terrorism or immigrants this week? Oh, just ID fraud, ok.)

For me, I'm naturally distrustworthy of technology, but that's only because I work with it every day, and relatively understand the problems and successes involved - especially when it comes to government implementations. My distrust has built up over time :) Actually explaining this, succinctly, to other people, is a different matter.

Without wishing to sound pompous et al, my own personal criticisms of the bill are also embedded in deep philosophical concerns, and finding a way to elucidate the role of individuals, governments and nations in a global state is probably even more difficult to get across. But it's worth trying, I think.