Nooface | Speech Recognition Becoming Key Security Function

Nooface

- Home
- About
- FAQ
- Discussions
- Journals
- Messages
- Topics
- Contact

- Preferences
- Older Stuff
- Past Polls
- Submit Story

Resources

- 3D User Interfaces
- Archives

Speech Recognition Becoming Key Security Function

posted by Editor on Monday December 23, @05:13PM

This article from the BBC News reports thats speech recognition is becoming a key security function in the post 9/11 era. Speech recognition technology was being dismissed as a "holy grail" when it failed to become part of everyday life after many promises. Now, products from vendors such as Nuance can be used to secure access to buildings using only a telephone. Users pick up a phone in the lobby and say who they are. The system will authenticate them using their voice, and buzz them into the building automatically. The users don't even have to say a specific password, and only have to talk to be authenticated. Moreover, it has to be a live voice, since a tape recording of the voice will not actually replicate that person's voiceprint, making it a secure system.

< Controlling Everyday Appliances With Handheld Computers | Microsoft As The 21st Century Utility >

Nooface Login

Don't have an account yet? Go Create One. A user account will let you customize the site's content according to your preferences. It will also allow you to moderate the comments of other users.

This discussion has been archived. No new comments can be posted.

Security and spoofing, and the tech arms race (Score:1)

by werejackal on Tuesday December 24, @12:57PM EST (#1)
(User #360 Info)

Moreover, it has to be a live voice, since a tape recording of the voice will not actually replicate that person's voiceprint, making it a secure system.

Ah, yes, but how do they guarantee that? This reminds me of that excellent article on biometric security devices in the german technology magazine (wish I had the link), which did a study about how easy it was to spoof fingerprint-based and face recognition-based biometric devices. A voice-based authentication technology will run into the same problem.

Why? Think about - /how/ will Nuance make sure it's a live person and not a recording? The face-recognition software tried to check for 'moving pixels', to prevent somebody holding up a photograph. This was easily circumvented by holding up a video image up to the sensor. The voice authentication technology will face a similar challenge.

If you think about it, there's not that many ways to make sure there's a live person on the other end. They usually boil down to questions of resolution and "flaws". If they're trying to authenticate by resolution, they're relying on the fact that a live voice will have some kinds of frequencies that a recording will not. Of course, the voice is transmitted through a phone line, anyhow, so some of those frequencies are lost. But you can see how this becomes a technology race - how sensitive is the security device vs. how well can a recording reproduce the voice.

Similarly, if they're relying on "flaws" to authenticate, like the face-recognition software did (i.e. make sure there's tiny imperceptible variations/movements), a good recording (or a good interference/whatever algorithm) can circumvent that.

I can think of only one way to do voice authentication that couldn't be thwarted by recordings - although it didn't sound like they're taking that path (their product description touts authentication by user saying their name). And that's by having the user say a random set of words, instead of a fixed name or password. The software would analyze the user's voice for common elements, and since the words to be spoken is different each time, no recording is possible.

It's an interesting problem, nevertheless. Always make sure to challenge vendor claims such as 'it /has/ to be a live voice'.

the UI of authentication. (Score:1)

by tchang on Monday January 13, @07:09PM EST (#2)
(User #25 Info) http://www.sonic.net/~tchang

Probably true that this could never be as fool-proof as they say. Of course there are also different levels of security. I, myself, have passwords with differing "secureness". For example, if I want to create an account for a website or something, I may choose a password that's guessable (for me or anyone else), certain email passwords may have higher security, other email passwords will have even higher security, and on from there.

Of course the UI is completely different. Voice-based authentication may be a great way to make the UI "invisible" for small-to-medium security levels. Especially if you're not required to mutter any specific "open sesame" password. Additionally, your voice isn't generally something that you lose, either, such as a key.

Anyway, I can imagine that there are other concerns too, such as if you have a cold, going through puberty, or just plain have a gag on and your voice changes or can't talk.

I'm not a robot like you. I don't like having disks crammed into me... unless they're Oreos, and then only in the mouth. -- Fry