SIR – THE 50TH QUEEN’S SPEECH delivered by Her Majesty Queen Elizabeth II in the House of Lords on 23 November was dominated by measures to combat terrorism, serious crime, drug abuse and anti-social behaviour.

Six separate Home Office Bills are being proposed to tackle these and related issues, with a further five Bills including at least some elements of law and order.

Arguably the most controversial of all the Bills, the Identity Card Bill sets out a national identity scheme comprising a national identity register together with individual cards for all UK citizens. The cards will be phased-in within the next four years, and could well be compulsory by 2011.

The Bill clearly states that biometric details will be used to identify and verify individuals, but how exactly? Biometrics aren’t perfect, of course. Some of them are pretty good, but is any one method good enough? A biometric that correctly identifies people 99.99% of the time could cause tens of thousands of exceptions per day (of which only a fraction may be attempts at benefit fraud or illegal immigration, etc). In such circumstances, real violations might be written off as systems failures and ignored.

Therefore, the UK’s impending national identity infrastructure may have to use more than one biometric (it would be “multi-modal”, to use current jargon) in order to achieve each of the Government’s stated goals for the cards – namely controlling immigration, fighting terrorism, reducing crime and improving public service delivery.

This doesn’t mean hundreds of biometrics and the Government recording everything from a person’s gait and body odour through to their weight and inside leg measurement on a huge database. What it does mean, though, is that the Government will have to think ahead and plan its strategy very carefully.

As the registration of biometrics is a complex process (because it must be very, very secure) and one that’s time-consuming, it’s also very expensive. Choosing how many biometrics there should be, and which biometrics they are, depends on balancing the cost of the system with its requirements.

There are really two separate kinds of requirement: identity and verification. The identity requirement is that a person’s biometrics uniquely identify them on the national identity register. The register will be used for ‘1:n’ matching: no two records on that register should contain the same set of biometrics, such that given a set of biometrics the register will return only one record (or no records at all if that person isn’t on the database).

The verification requirement is quite different: given an identity card and a person, the card can check whether a person is its rightful owner. This involves a ‘1:1’ match. Given a biometric, the identity card can say ‘Yes’ or ‘No’.

Which biometrics should be used to satisfy these requirements, then? A good system design policy is that they’re different. One might argue, for example, that none of the biometrics stored in the national identity register can be non-invasive. In other words, a determined hacker shouldn’t be able to take someone’s picture without their knowledge and then look up that picture in the national identity register to find out who they are.

In truth, only invasive biometrics should be used for that purpose, so that when a policeman requires you to place your finger on a reader, or place your eye up against an iris scanner, you know that you’ll be identified.

Conversely, one might suggest that only non-invasive biometrics be used for verification so that an individual might walk up to the door of a Government building and wave their card. A camera may then take a picture of the person’s face and send it to the card for matching. If the card says: “Yes. This is my owner” and the door says: “Yes. The owner of this card is allowed in” then the door would open.

Without any reference to the national register.

If the system is going to deploy non-invasive biometrics for verification purposes, what should they be? There are a number of candidates: facial recognition and voice recognition, for example.

Just as it’s wrong for some to claim that biometrics are a ‘magic bullet’ that will solve all of our identification and verification issues, it’s equally wrong to say that biometrics aren’t the answer.

In fact, biometric technologies are an extremely good answer. As long as you’re asking the right question.

Neil McEvoy, Managing Director, Consult Hyperion

Source

SMT