Monday 17 October 2011
Author: Dan Jellinek
India builds world's largest biometric ID database
Details of how some of the world's most underprivileged people could be brought government services for the first time with the help of biometric identification technologies are to emerge at this year's Biometrics 2011 conference, opening in Westminster tomorrow.
Facial recognition systems using high definition cameras to identify people in controlled settings such as a border post or informal situations such as a crowd on a railway station platform, are underpinning a universal citizen ID system being rolled out across India's 1.2 billion population.
The Unique Identification Authority of India's (UIDAI) Aadhaar program - taken from the Sanskrit word for base or foundation - will create the world's largest database of individuals at an estimated cost of up to US$4.4bn. It will use multiple types of biometric data including retina scans, fingerprints and multiple facial images. A consortium consisting of Accenture with software firm Daon and Indian outsourcing specialist MindTree was one of three technology groups selected last year to build the system.
Mark Crego, senior defence and public safety executive at Accenture, told UKAuthority.com the fingerprint and iris data has already been captured for 100 million people, with facial recognition as a further check that data is correct by matching numbers with gender and age band. "Sometimes you might be enrolling the whole family, in which case the officer entering the data might get it wrong. But using the technology, you can match on age as well, so the system could see if a male adult number was being wrongly ascribed to a 12-year-old girl, for example.
The vast process of registering all India's citizens, including those in extremely remote areas, inevitably involves a mix of high and low tech solutions, Crego said. While registration officials would have access to high-tech portable or mobile enrolment stations, often the results of registration including biometrics would have to be saved onto encrypted thumb-drives and sent back to base in the ordinary postal service, he said.
Although face-matching technologies have been in use for more than a decade, huge improvements to the algorithms used have led to far greater accuracy over the past five years, Crego said. "What we see now has nearly the same accuracy as two fingerprints, in controlled circumstances."
While facial recognition is still best-known for use at national borders - such as at Heathrow and Stansted airports - they can also now identify travellers even before they get to the gate, for example as they leave a train at the airport, he said. Systems can also be used to check that people in a controlled area of the airport such as runway or hangar areas are supposed to be there. They can also recognise gender and estimate age of a group or crowd of people, Crego said, which could be of use in situations such as sporting events to identify where security personnel might be needed most without slowing down the flow of people into an area with formal individual checking.
Details of how some of the world's most underprivileged people could be brought government services for the first time with the help of biometric identification technologies are to emerge at this year's Biometrics 2011 conference, opening in Westminster tomorrow.
Facial recognition systems using high definition cameras to identify people in controlled settings such as a border post or informal situations such as a crowd on a railway station platform, are underpinning a universal citizen ID system being rolled out across India's 1.2 billion population.
The Unique Identification Authority of India's (UIDAI) Aadhaar program - taken from the Sanskrit word for base or foundation - will create the world's largest database of individuals at an estimated cost of up to US$4.4bn. It will use multiple types of biometric data including retina scans, fingerprints and multiple facial images. A consortium consisting of Accenture with software firm Daon and Indian outsourcing specialist MindTree was one of three technology groups selected last year to build the system.
Mark Crego, senior defence and public safety executive at Accenture, told UKAuthority.com the fingerprint and iris data has already been captured for 100 million people, with facial recognition as a further check that data is correct by matching numbers with gender and age band. "Sometimes you might be enrolling the whole family, in which case the officer entering the data might get it wrong. But using the technology, you can match on age as well, so the system could see if a male adult number was being wrongly ascribed to a 12-year-old girl, for example.
The vast process of registering all India's citizens, including those in extremely remote areas, inevitably involves a mix of high and low tech solutions, Crego said. While registration officials would have access to high-tech portable or mobile enrolment stations, often the results of registration including biometrics would have to be saved onto encrypted thumb-drives and sent back to base in the ordinary postal service, he said.
Although face-matching technologies have been in use for more than a decade, huge improvements to the algorithms used have led to far greater accuracy over the past five years, Crego said. "What we see now has nearly the same accuracy as two fingerprints, in controlled circumstances."
While facial recognition is still best-known for use at national borders - such as at Heathrow and Stansted airports - they can also now identify travellers even before they get to the gate, for example as they leave a train at the airport, he said. Systems can also be used to check that people in a controlled area of the airport such as runway or hangar areas are supposed to be there. They can also recognise gender and estimate age of a group or crowd of people, Crego said, which could be of use in situations such as sporting events to identify where security personnel might be needed most without slowing down the flow of people into an area with formal individual checking.
Despite all the recent improvements in accuracy, however, even the best facial recognition systems are still wrong with false matches or incorrect clearances up to 5% of the time, Crego said. This means the technique must be carefully managed alongside other data to ensure "you don't cause officers to jump when they don't need to jump."
The Unique Identification Authority of India http://uidai.gov.in/
The Unique Identification Authority of India http://uidai.gov.in/
