Mind reader

By the end of 2005 the first biometric-enabled passports will be issued within the UK. Containing a biometric facial recognition image, the document will act as a precursor to the introduction of a compulsory national identity card, to be phased in from 2008.

The government hopes that a biometric facial image will help to counter identity fraud and verify the identity of the holder against the document. And if present policy is continued, along with facialinformation the card is also likely to contain both fingerprint and iris scan data.

Behind all this is the belief that such technologies are now mature enough to work in a national system. But according to Angela Sasse, Professor of Human-Centred Technology in the department of computer science at University College, London, much work is needed to match the reality of human behaviour with the performance expected from the technology.

She has spent years studying human responses to biometric technology and says that unless more research is carried out into its limitations users may become unco-operative and any systems implemented may fail.

Sasse read psychology in her native Germany before moving to the UK, where she gained an MSc in occupational psychology from Sheffield University, followed by a PhD in computer science from the University of Birmingham.

She then worked as a human factors specialist for Philips Corporate Industrial Design before joining UCL’s department of computer science in 1990. Since 1997 her research has focused on the human element of security systems.

As a specialist in this field, she contributed to the EU’s Biovision project, which produced a roadmap for the use of biometrics in Europe until 2010. Last year she completed a study concerning the usability and acceptance of biometrics as part of a field test commissioned by the German Federal Office for Information Security.

She has also acted as a specialist adviser to the Home Affairs Committee for its inquiry into the introduction of ID cards, providing information on practical aspects of implementation - or, put simply, how the public is likely to behave when faced with them.

According to Sasse, the implementation of a national biometric-based system depends on success in three areas: universal access, or the ability of every citizen to join the scheme, enrolment and verification.

‘If the person has arthritis, long fingernails, is wearing handcream or has circulation problems then this will not produce a good reading,’ said Sasse. ‘Fingers can be cut or bruised, either accidentally or deliberately to prevent enrolment, or fingerprints can be worn down by age, chemicals or manual labour. Groups such as Asian females have naturally very fine prints that are hard to register.’

Iris scans are also problematic. ‘Around half the population wears glasses or contact lenses and the population in the West is ageing, so a machine’s ability to read through them is essential,’ she said, pointing out that if the wearer is asked to remove them they may not be able to see to line up with the reader.

‘In tests people got confused and presented the wrong part of their face. They had problems focusing and needed a mirror for feedback, otherwise they ended up presenting their nose to the machine. If the camera is able to give feedback it must be instant so the user does not move in the wrong way.’

Finding equipment that can be adjusted for wheelchair users and those measuring under 1.55m or over 2.1m can be a problem too.

But even users of average height may struggle. ‘We found that people were reluctant to touch the equipment, even when instructions to do so were present,’ said Sasse.

‘Part of this is for reasons of hygiene. Rationally there’s no more bacteria on a fingerprint scanner than there would have been on the door handle of the room. However, the light beneath the scanner means you can see any dirt. As the reader is warm people start to imagine bacteria breeding.’

Unfortunately, this has implications for using the system for immigration. ‘In an airport, with lots of people from different backgrounds and nationalities using it, it would bring out social fears, however unacceptable.’

Holding the required neutral expression during verification can be difficult too, depending on where the system is to be used. ‘At Gatwick when you are going on your holidays and have had a few pints in the airport bar nothing will crack people up faster than demanding this [posing for a facial reader],’ said Sasse. ‘If the system is slow then people also get bored and do things such as pull faces or start to look away.’

SmartGate, a facial recognition system trialled by Customs at Sydney International Airport since November 2002, is designed to confirm identity of Qantas aircrew. Hailed as a success, the system uses five high-quality images to compare against the person presenting at passport control. Average processing time is around 17 seconds per person.

So far results have shown that the percentage of users incorrectly rejected was two per cent. However, Sasse said achieving such results has taken much work. ‘They had to change the airport lighting, train staff in how to position themselves for a reading and move to live updating of images, recalculating the facial template held by the system each time the person goes through to account for changes of hairstyle and so forth,’ she said.

‘In a national system this constant change would create a huge security problem and enormous costs owing to the huge amount of stored data.’

Another option would be to use a template system based on measuring the distances between set facial points. ‘To ensure interoperability between countries, the same equipment and software would have to be used worldwide,’ said Sasse. ‘This is not going to happen.’

The amount of time taken by the system to take a reading and post a result is vital in determining whether or not it can carry out the task for which it was designed. There is nothing the public hates more than queuing, especially at airports.

Sasse said that while the reading and matching times of some manufacturers look impressive, they do not factor in human behaviour. This includes the need for the user to put down their bags, find their card and then begin the process.

With this factored in, an average usage time of 12-20 seconds could be achieved by a frequent user, as proved by a UK trial that took place last year at the Venerable Bede School in Ryhope, Sunderland. The school tested iris scanning to identify pupils entitled to free meals. However, the project was scrapped for being too slow, managing to achieve a scan rate of five pupils per minute rather than the target of 12. School officials complained that children’s meals were going cold in the queue.

‘At UCL, after our tests we thought a throughput rate of five people per minute was actually pretty good compared to the results we had seen,’ Sasse said.

Recently, with the passport implementation looming, she has been part of a coalition calling on the government to examine issues beyond whether or not the basic technology works. ‘We have called for a scientific commission to be set up to define what work needs to be done, but nothing has happened,’ she said.

‘At the end of the day the success of schemes depends upon how many of the benefits being proposed will be felt by users and how the financial and social cost of the scheme will affect people’s relationship with the state.’