Speech Technology Magazine SpeechTEK Conference
 

Another Layer of Security

STM Blog @ 4:17 pm

Opus Research’s third annual Voice Biometrics Conference in NYC today.  As we’re actually preparing to ship the June issue, I’m not live blogging from it, so I can’t give a thorough convention report.

Voice biometrics strikes me as an area that people have a great deal of hope for, even if there aren’t a great deal of publicized consumer-facing deployments (especially in the US).  Still, Opus senior analyst Dan Miller thinks the market will pretty huge by 2011.

My understanding of it is that while the technology is there, vendors are still experimenting with how best to deploy it. Additionally, enterprise customers, as always, are a little hesitant to adapt this new technology.

For instance: End customers still don’t know what exactly the technology is. To the extent that the public thinks about biometrics at all, Hollywood has given them images of retinal scans (see: Demolition Man, Minority Report) or fingerprint scans (see: Men in Black, Bourne Ultimatum–though, ok, that movie did showcase voice biometrics being compromised).

So how do enterprises educate customers about voice biometrics?

Julia Webb, VoiceVault’s EVP of sales and marketing said user involvement is critical: enterprises should explain the benefits of voice biometrics to callers during the enrollment process, even at the expense of longer call durations.

In fact, Rex Stringham, president and CEO of EIG, said during a test, callers preferred a longer enrollment process so long as the benefits and the technology were explained. (He also said that the sampling was “sizable” though not statistically significant).

Stringham also said that immediately bombarding the caller with detailed explanations tended to lead to refusals to enroll. Start the dialogue, he said, assuming that no help is needed: “Assume everyone’s a power user until they prove you wrong.”

Miller pointed out that none of the voice biometrics panelists agreed on how to enroll a customer into the system. Problem: how do you make sure that the person whose voice print you’re collecting is in fact the person he claims to be?

Stringham conceded that voice biometrics offered a “semi-secure” password in that “once it’s established, it’s secure.”

Studies conducted by both Stringham and by Harris Interactive (on behalf of Nuance) indicated that customers wanted a combination of security and convenience. Occasionally, these can be mutually exclusive. So if the enrollment process is too convoluted, customers will opt out. Of course, if the enrollment process itself isn’t secure…well, that pretty much undermines voice biometrics as a useful layer of security.

How to reconcile this?  I don’t think there’s any agreement yet. Webb mentioned two extremes: on the one hand, you can send a password through the mail. On the other hand the FIPS 201 standard actually has people physically show up to register. They’re fingerprinted and a square inch of flesh off their back is hacked off and preserved for future reference (kidding about that last part).

Anyway, voice biometrics companies will have to come up with an effective middle ground.

No Comments »

No comments yet.

RSS feed for comments on this post. TrackBack URL

Leave a comment

Previous Posts
Keyword Tags
Archives
© 2008 Speech Technology Media, a division of Information Today, Inc. About/Contacts | PRIVACY POLICY