Collar the lot of us! The biometric delusion
Optimism beats evidence in the drive to fingerprint the world
Faced with revolution, the government would have to abandon the NIS. Logic, maths, science, a basic understanding of technology, businesslike common sense, an adult sense of responsibility and simple truth-telling all suggest that the NIS should have been abandoned on the day the biometrics enrolment trial report was published*.
Instead, what did IPS say when the House of Commons Science and Technology Committee confronted them with these failure rates? According to the Committee’s July 2006 report, IPS said that the key findings of their biometrics enrolment trial were not key findings, and that the trial was not a test of the reliability of biometrics, but only a test of their usability (para.88).
If it wasn’t a test of reliability, why are the reliability figures reported as key findings? Why would IPS want to test usability but not reliability? Surely they wouldn’t deploy the NIS with biometrics that are congenial to everyone but just don't happen to work. And what is this distinction between reliability and usability? For 300 pages, the May 2005 report discusses usability almost entirely in terms of reliability.
Despite the polite and sensible entreaties of the Committee, no large-scale field trial of the reliability of flat print fingerprinting has been subsequently conducted by IPS. If the biometrics enrolment trial was not a reliability test, then there is still no evidence to support IPS’s claim that flat print fingerprinting can deliver their vainglorious ambition.
According to the House of Commons Science and Technology Committee report, "on 6 March 2006, we met informally a group of senior policy advisers from the Department of Homeland Security to discuss the identity cards programme. When questioned about the maturity of biometric technologies, the advisers agreed that currently the technology was probably not as reliable or as accurate as it might need to be for a national identity card scheme" (para.81).
Logic, maths, science, etc… all having been abandoned, IPS told the Committee, not quite that the earth is flat, but that the maximum acceptable false non-match rate for flat print fingerprinting is one per cent (para.18) and they pointed the Committee (p.126ff) to a May 2004 report written by the US National Institute of Standards and Technology (NIST).
* The false non-match rate associated with IPS's chosen biometrics varies between 19 and 52 per cent. Really? Is that true? There is an obvious counter-example – schools up and down the country use biometrics to take the register, to manage library-lending and to operate cashless canteens. Why don’t they suffer from 19-52 per cent false non-match rates? The answer is given in NIST's May 2004 report. Please see section 4.3, Trading FRR for FAR, pp18-19. (FRR = false reject rate = false non-match rate. FAR = false accept rate = false match rate.) Schools can calibrate their biometric equipment to operate close to a zero false match rate or close to a zero false non-match rate, one or the other, but not both. There is a trade-off. If the school goes for a low false match rate, they will inevitably get a high false non-match rate and vice versa. They go for a low false non-match rate so that not too many pupils starve. That's why there is no false non-matching problem to report.
But there must be a concomitant false matching problem. According to Figure 9 of the NIST report (pp.16-17), with a low false non-match rate, the false match rate can quickly rise to 10 per cent and even higher. Suppose that a pupil collects his or her lunch from the canteen and is identified only by his or her flat print fingerprint. Then, in a school of 1,000 pupils, it is likely that the computerised biometric canteen system can't identify which of 100 pupils in the school is the luncher and that the school is therefore wasting its money.
The implication is instructive – it is that biometric identity is discretionary. Depending on how the operator calibrates the equipment, the biometrics might say that you are you, or they might not. That is not how we usually understand identity. IPS is using an alien version of the concept of identity.
Sponsored: DevOps and continuous delivery