X
Business

Feds study voice authentication

The federal government is commissioning a trial of voice authentication technology to see whether it is mature enough for use by its agencies.The University of Canberra is set to begin the trial on behalf of the government following the donation of two servers by Unisys Australia.
Written by Kristyn Maslog-Levis, Contributor
The federal government is commissioning a trial of voice authentication technology to see whether it is mature enough for use by its agencies.

The University of Canberra is set to begin the trial on behalf of the government following the donation of two servers by Unisys Australia. "As government departments frequently exchange security information over the telephone, it is of paramount importance that they can trust that the voice at the other end of the line is who they say they are," Clive Summerfield, adjunct professor for computing and head of the University's National Centre for Biometrics, said in a statement.

He said the centre was evaluating how effective a range of vendors' offerings were in identifying voices on fixed or mobile phones.

"We are giving them all the same test to do in order to see how well they do on it. We give them all type of speech data, mixing some good data with spoiled data in an attempt to confuse the technology. For instance, we add background noise to the voice data we give them for testing," Summerfield said.

Voice authentication uses voiceprints and pattern recognition software to verify a speaker. It relies on the fact that vocal characteristics, like fingerprints and the patterns of people's irises, are unique for each individual.

Voice authentication vendors such as US-based companies Nuance and Scansoft provide voiceprinting software to various industries worldwide. The Nuance Verifier uses individual voiceprints to deliver high security without the use of passwords and PINs.

Summerfield, however, declined to comment on which vendors and which offerings were involved in the evaluations.

The professor told ZDNet Australia  he could not disclose which government departments were interested in the technology, but added that the federal government is the "prime mover" in getting the research done.

He said existing authentication systems were inadequate. "How do you confirm the identity of a caller? At the moment the process requires asking you 20 questions, but just because someone knows the address, telephone number and name doesn't mean that is actually the person. The process itself is flawed in many ways and at the same time it is an expensive process to adminster, essentially chewing up call centre time," he said.

Summerfield said banking and other industries in the commercial sector were also interested with voice authentication -- particularly with reference to its application for call centres. Customers can speak to automated services that verify their identity from their unique speech characteristics and associate this with their entitlement to the service.

Unisys donated two ES3120 servers for the duration of the project, one of which runs Microsoft Windows 2000 and the other Linux. The professor said the servers were "fundamental" in the running of the large speech database during the trial.

"The high specs capability of the servers act as fast and substantial database service. We also have to capture the results that the evaluations are giving us and this needs to to be stored in a big database in order to be able to cross match and cross check it. Without these servers, it would be much more problematic in a logistics point of view," Summerfield said.

Although he could not disclose the budget for the undertaking, Summerfield said it was "the most expensive evaluation yet undertaken". It is also said to be the largest test of its type ever conducted anywhere in the world.

Summerfield believes voice authentication is second to iris recognition and higher than fingerprint authentication in terms of security efficiency.

Editorial standards