Sponsored By

Listen to My Voice, Tell Me Who I AmListen to My Voice, Tell Me Who I Am

Voice biometrics could be the future of personal verification, but are consumers ready?

Gary Audin

April 13, 2018

3 Min Read
No Jitter logo in a gray background | No Jitter

Can I identify you using voice recognition technologies? There are two parts to this question. First, does the technology actually work effectively enough so that the recognition is accurate? The second part is, do callers believe it works well and can be depended upon to be consistently viable to verify them?

Does Speaker Recognition Work?
I have worked with speaker and speech recognition many times in my career. I've learned that recognizing who a person is by their voice is different than recognizing the words they speak. I primarily work with Nuance's Dragon speech-to-text conversion program; it's accurate, but not always. Of the demonstrations of speaker recognition I've seen at exhibits and conferences, most seem to have accuracy in the range of 85% to 90%.

As I learned in a recent Pindrop Blog post, biometric technology offers a newer way to combat fraudsters by enabling identification through something a person is -- like retinas, fingerprints, and speech -- rather than something a person knows -- like passwords. But when it comes to voice biometrics tools, their success is often determined by the quality of the underlying machine learning tools. In other words, biometrics will work for identification, but it's not guaranteed to be perfect. For example, synthetic voice can be used to falsifying a speaker's voice, thereby circumventing speaker recognition/voice biometric technologies.

portable

Voice Recognition Acceptance
How do consumers view voice recognition? According to a Pindrop study conducted by Harris Poll of more than 3,000 U.S. adults, nearly half of Americans indicated they would be likely to use voice recognition as a form of personal verification. Further, 81% of respondents believe there are benefits to using voice recognition for the purposes of personal verification.

Customers Drive Voice Verification
Voice recognition could be particularly useful for contact center scenarios. The Pindrop study found that among consumers' top complaints are the number of steps they must go through to prove who they are when calling into a contact center for support or assistance. Furthermore, 18% of respondents indicated that that experience is difficult because they were unable to remember the answers to their own security questions and were thus locked out of their accounts. Voice/speaker verification can go a long way toward reducing these frustrations.

Uncertainty with Speaker Recognition
Even if 81% of respondents expressed a belief in the benefits of voice recognition for verification, concerns still exist. On the other hand, 94% reported that they think there are drawbacks to this technology. The primary reasons for skepticism include:

  • Background noise may cause the tech to not work well (61%)

  • Accuracy of the technology (60%)

  • As mentioned above, there is concern about voice cloning/synthetic voice that can fool the recognition system (48%)

  • Accents may prevent widespread automated voice recognition (43%)

  • Security limitations (39%)

  • If the user does not speak loudly and clearly, while in a quiet environment, the technology will not work (36%)

  • Because this is a new technology, many people will not know how the verification mechanism works and therefore may distrust it (31%)

Voice is not going to disappear in our economy. Voice activated services are increasing; take, for example, the products and services from Google and Apple. There are UC&C platforms that can be voice activated. There is an effort to increase security through multifactor authentication, with voice being one of those factors. Unfortunately, as voice activated mechanisms become common, fraudsters will leverage other technologies to bypass the security.

We need to employ speaker recognition. We also need to believe it works. More reports on the security effectiveness of speaker recognition need to be performed.

Related content:

About the Author

Gary Audin

Gary Audin is the President of Delphi, Inc. He has more than 40 years of computer, communications and security experience. He has planned, designed, specified, implemented and operated data, LAN and telephone networks. These have included local area, national and international networks as well as VoIP and IP convergent networks in the U.S., Canada, Europe, Australia, Asia and Caribbean. He has advised domestic and international venture capital and investment bankers in communications, VoIP, and microprocessor technologies.

For 30+ years, Gary has been an independent communications and security consultant. Beginning his career in the USAF as an R&D officer in military intelligence and data communications, Gary was decorated for his accomplishments in these areas.

Mr. Audin has been published extensively in the Business Communications Review, ACUTA Journal, Computer Weekly, Telecom Reseller, Data Communications Magazine, Infosystems, Computerworld, Computer Business News, Auerbach Publications and other magazines. He has been Keynote speaker at many user conferences and delivered many webcasts on VoIP and IP communications technologies from 2004 through 2009. He is a founder of the ANSI X.9 committee, a senior member of the IEEE, and is on the steering committee for the VoiceCon conference. Most of his articles can be found on www.webtorials.com and www.acuta.org. In addition to www.nojitter.com, he publishes technical tips at www.Searchvoip.com.