U.K. agency: Iris recognition needs work

Speaking at the Biometrics 2005 conference in London on Thursday, Rob Bowley, director of ID Projects with the UKPS, said the current technology must be improved to carry out more efficient biometric scanning. "We've got a lot of work to do--the technology has got to work in all environments. We will be looking at various types of iris cameras," he said.

The UKPS carried out the first large-scale U.K. trial of the three main biometric techniques between April 14 and Dec. 24, 2004.

The trials, which took place in various government offices in London, Newcastle, Leicester and Glasgow, involved simulating the enrollment process for entering biometric details into a central database. Researchers from the UKPS measured the amount of time taken to process about 10,000 participants and how customers reacted to the technology.

Participants in the trial tested three biometric devices, including facial-, iris- and fingerprint-scanning technologies. Individual enrollment attempts took eight minutes and 15 seconds on average. To accurately simulate a large-scale deployment, researchers preloaded the databases with 118,000 irises and 1 million fingerprints. "It wasn't a trial of the technology; it was a trial of the way technology interfaced with the people," Bowley said.

For iris enrollment--the term used to describe the process of registering an iris image on a system--the success rate for the 10,000 participants was 90 percent and 60 percent for able-bodied and disabled users respectively, compared with the 100 percent and 96 percent for fingerprint recognition.

The UKPS said that the trials revealed that much of the current iris-scanning technology is not flexible enough to cope with the demands of dealing with a variable public. In particular, participants with visual impairments and above-average height could not access the system easily and sometimes could not access it at all.

However, Bowley admitted that a lack of skilled operating staff was also a cause for the difficulties with iris biometrics. "You can have all the best equipment in the world, but if you have badly skilled staff operating the systems you will still get bad results," he said.

If iris recognition is going to be widely deployed, then manufacturers must create flexible technology that adapts to the variation among individuals rather than the current one-size-fits-all system, said Bowley. "We need systems that can be adjusted for all the different types of people. Very tall people and those in wheel chairs suffered problems with accessing the machine," he said.

Karen Gomm of ZDNet UK reported from London.