With a growing number of
financial institutions deploying the technology and conducting
extensive studies and pilot schemes, voice biometrics finally
appears set to fulfil its potential as a fraud reduction mechanism,
writes Paul Golden.
Go deeper with GlobalData
Biometrics is a
collective term for a number of different technologies that can be
used to automatically confirm a person’s identity. They do this by
comparing their physical or behavioural characteristics in real
time against enrolled computer records.
The number of uses it is being put
to is, staggering – from airport security to transaction
verification.
Biometric capture devices create
electronic digital templates that are encrypted, stored and then
compared to encrypted templates derived from ‘live’ video or
audio.
These templates are generated from
complex and proprietary algorithms and encrypted to secure and
protect them from disclosure. On their own, biometric templates
cannot be reconstructed, decrypted, reverse-engineered or otherwise
manipulated to reveal a person’s identity.
How well do you really know your competitors?
Access the most comprehensive Company Profiles on the market, powered by GlobalData. Save hours of research. Gain competitive edge.
Thank you!
Your download email will arrive shortly
Not ready to buy yet? Download a free sample
We are confident about the unique quality of our Company Profiles. However, we want you to make the most beneficial decision for your business, so we offer a free sample that you can download by submitting the below form
By GlobalDataVoice recognition systems use the
person’s voice as the password by comparing its tone and timbre
against a database entry. Joining or enrolling in a voice
recognition system may take a little time – since the user must
‘teach’ the computer by speaking a number of phrases – but after
this initial phase the technique should be capable of being used
quickly and efficiently.
Deployment
delays
The technology
required to support such systems has been around for a while, but
concerns over instability have held back large-scale
deployment.
The Ovum report, The Future of
Voice Biometrics in the Enterprise Market, published in late
2009, stated that, despite marked improvements and the ability to
create robust solutions, voice biometrics still has not enjoyed the
widespread proliferation purveyors of the technology
anticipated.
However, the report also pointed to
an upturn in the market supported by stability in the financial
services sector, one of the major market prospects for voice
biometrics.
This view is supported by a report
published in November last year by BCC Research (Biometrics:
Technologies and Global Markets), which predicted sustained
growth in the global market for various biometric technologies over
the next four years.
According to BCC Research, voice
recognition biometric technologies (together with face, iris and
vein systems) form the second largest segment of the biometrics
market and will grow in value from $1.4bn in 2010 to $3.5bn by
2015, a compound annual growth rate of 20%.
Given that mobile devices come with
access to the biometric sensor required for voice verification in
the shape of a microphone, it is perhaps surprising that voice
biometrics does not already play a leading role in the battle
against payment card fraud.
According to Adam Shearn, head of
enterprise pre-sales at speech and imaging solutions provider
Nuance, this can partly be explained by card companies’ fears they
will be targeted by every hacker who wants to make a name for
himself.
“However, the experience of clients
to date proves this is not the case and we expect to see a number
of new deployments in the UK financial services sector this year,”
Shearn says.
Sean Keane, general manager (EMEA)
for customer communication at solutions provider Salmat, believes
voice biometrics solutions are well placed to play a major role in
card fraud reduction.
“Harnessing biometric technology
can deliver better authentication results than traditional
knowledge-based methods, as it confirms the ownership rather than
just the possession of security credentials,” Keane says.
“For ‘card not present’ fraud,
voice biometrics is one of the most secure means of verifying
someone’s identity remotely and it is simple to deploy as the
infrastructure is already in place – everyone has access to a
telephone and their own voice.
“Technologies such as card readers
can be effective, but are costly and complex to deploy as they rely
on the distribution and maintenance of specialised hardware.”
Off the ground
Voice biometrics is
gaining traction in Australia with organisations such as Citibank
(which started trials last autumn) and National Australia Bank.
Keane says a number of UK card
issuers have initiated voice biometrics studies and pilots over the
past year.
He adds: “Institutions [in this
country] will have noted how being first to market with such a
genuinely powerful security enhancement can be marketed as a real
and compelling differentiator to the public and, interestingly, as
a deterrent to professional fraudsters, who often profile and
pursue those institutions perceived to be less secure”.
Once customers start to use voice
biometrics to access services such as banking, they will quickly
expect the same kind of security from their other service
providers, says Peter Galloway, head of voice self-service at
contact centre solutions Sabio.
“Biometrics has potential value in
lots of other ways based on passive identification of callers,”
Galloway says.
“One example could be to support
fraud hot-spotting, with biometrics spotting the voiceprints of
users who have already demonstrated a propensity towards fraud and
immediately advising agents that the caller may be fraudulent.”
But while the technology is mature
and systems are available, Galloway admits there are a number of
constraints to deployment. One issue is creating an effective
enrolment strategy where the customer is in the system and another
is about convincing compliance-side people with real data relating
to the accuracy and performance of the systems. A further
impediment was the credit crunch and the limitations it has placed
on capital investment.
Enrolment is a massive undertaking,
If customers were asked to comply within a certain period it would
inevitably alienate some users.
As a result, many vendors suggest
organisations adopt an opt-out approach even though this makes it
harder to accurately predict how many customers will use the system
and in what time frame.
Prime reference sites for this type
of technology are limited in terms of financial applications, but
Galloway is confident that will change soon with a major US credit
card issuer expected to launch voice biometrics in April.
“We believe this will create
massive new interest,” he says.
“The big banks are not yet ready to
proceed [the implementation cycle is closer to 24 months than 12
months], but had the credit crunch happened a year later voice
biometrics would already be a mainstream application in the
UK.”
Concerns and
challenges
Concerns over the
possible use of recorded voice prints by fraudsters are misplaced,
suggests RJ Langley, vice-president major systems at biometric
identity management company Daon.
“Any biometric or security system
can be ‘spoofed’ in some manner and different voice algorithms vary
in their susceptibility to spoofing,” Langley says.
“However, this type of attack
depends on the attacker knowing in advance which words or phrases
to pre-record.”
The system can also be set up to
ask the customer to repeat random information.
There is widespread agreement that
increased deployment of voice biometrics will not spell the end of
other authentication processes.
For example, Daon believes all
authentication mechanisms need to be multi-factor, explains
Langley.
“Relying on a single factor ignores
the fact that in client bases ofms of people, some are going to
have trouble with that single factor – for instance, what does a
person do with a voice system when they have laryngitis?” he
asks.
“Then we have to remember that
biometrics are not perfect – they occasionally fail to correctly
identify the person in question. There are varying levels of
security requirements and in those instances where greater security
is needed, multiple factors are the logical approach.”
Keane agrees that attempting
biometric verification on a poor-quality phone line or in a very
noisy environment decreases the likelihood of obtaining accurate
results, but adds that this is also the case where the call is
placed by a human.
He adds that, as in any
technology-dependent scenario, it is important to have contingency
plans in place for automation failures.
User
authentication
Voice biometric systems are more
effective when used to verify that a specific individual is who
they say they are, rather than being used to compare a random voice
print against a database of thousands of samples, explains
Galloway.
Biometric systems have a higher
tolerance than voice recognition and some systems will raise a
fraud warning if the voice print is too close to the recorded voice
print, which is known as ‘anti-aliasing’, Galloway adds.
But if the system is struggling to
get the audio it can be difficult to get a match. Some people have
predominant attributes that are more or less tolerant of noise, for
example someone who is softly spoken.
“The biometric engine is tuned to
deal with mobile phone acoustics, but background noise will have an
impact so we implement various error handling routines that prompt
the system to ask the customer to move somewhere quieter or speak
up,” says Nuance’s Shearn.
He expects some companies to rely
solely on voice prints for low value transactions and employ
two-factor authentication for higher value transactions or where
the sample falls below a certain ‘quality’ level.
Authentication is always going to
be multi-factor because you will never be able to say that one
piece of information is categorically accurate, concludes
Galloway.
“However, existing information can be combined with voice
biometrics to confirm identity and reduce the burden on customers
to remember multiple passwords and PINs,” he adds.
