All biometric systems are probabilistic

(12/11/2010)

"For nearly 50 years, the promise of biometrics has outpaced the application of the technology," said Joseph N. Pato, chair of the committee that wrote a recent National Research Council report. "While some biometric systems can be effective for specific tasks, they are not nearly as infallible as their depiction in popular culture might suggest. Bolstering the science is essential to gain a complete understanding of the strengths and limitations of these systems."

However, Stewart Hefferman, the CEO of OmniPerception, said that this report by the National Research Council that questions the accuracy of biometric technology is itself inaccurate and misleading. Mr Hefferman, who heads up the UK-based company, defended the accuracy of biometric technology following the report by the National Research Council.

"The report that says biometrics are not accurate or sometimes “fail” is itself inaccurate and misses the point," said Mr Hefferman, who has worked in the biometric industry for more than 12 years.

"Biometrics – whether we are discussing face, finger, iris, hand geometry or indeed any other biometric is, probabilistic. There is nothing new there. It always has been and always will be. There is also nothing new about the fact that biometric systems sometimes fail to correctly recognise someone. Lastly, there is nothing new about the fact that people may – very infrequently – be misidentified."

Mr Hefferman pointed out that problems or failures happen “very, very infrequently”.
"The fact that any biometric is probabilistic means that a good quality system built around any biometric product should recognise this," he argued.

"Frequently, biometric components within a system are blamed for the failure of the system because designers simply don’t understand – or want to recognise – the fact that all biometrics are probabilistic. Rather than sticking their heads in the sand and keeping their fingers crossed, they should design with this fact in mind."

Mr Hefferman said that to some extent it does not matter whether the error rate is 1% or 0.0000001%, it is still not zero and so needs to be addressed through appropriate design.

"The point is to design in the probabilistic nature of the technology. Make sure you know how to cope when people can’t enrol. Make sure you understand the impact on your system if someone fails to be correctly recognised and then – you never know, the solution might be as simple as even putting a microphone or camera next to the reader."

Related topics:  Eye biometrics   Face biometrics   Fingerprint biometrics   Hand biometrics   Other biometric systems   Voice biometrics 

Print version | Email to a friend | Related articles

Data breaches: Trends, costs and best practices gives you all the latest information on securing personal and corporate data, key recommendations for immediate action to improve data security, and how to respond to data breaches.

Other Security news and resources