Voice-based “captcha”?
April 23, 2008
Was talking with a friend last week who suggested the idea of voice-base captcha services. The main idea is that you’d record some basic words up front, then when prompted to say a word or collection of letters to authenticate, you’d say them to the site. The authentication system would do a voice analysis between the stored words and the new words. What do you think?
Did you like this post? Buy me a hot chocolate!
Posted in Ideas
April 23rd, 2008 at 12:30 pm
My question is: is this voice based captcha or voice based authentication. If I’m a user, I pre-record some words and then am instructed to say one, to me that is part of a multi part authentication system.
One of the problems with doing this as a CAPTCHA system is that there are then a limited number of possibilities, so it is easily broken. I would compare it to being required to recognize a picture of something. Since there are a finite number of pictures (or in this case spoken words) one could create a database with them all, compare the picture to one they have already seen and identify through the database.
Maybe I am misunderstanding what you are suggesting.
April 23rd, 2008 at 1:32 pm
More voice-based auth, but perhaps using the ‘captcha’ type interface.
If I record me saying ‘foo’ and ‘bar’ and ‘baz’ and whatnot, a system should be able to come up with a ‘voiceprint’ of how I say certain words. If I’m then asked to say some other words, a comparison should be able to determine if I’ve got the same voiceprint.
It’s just another biometric sort of thing, but perhaps voice is too open to imitation?? Also, I don’t want to be saying things out loud all the time. I’m the one who was just bitching about voice phone trees some time ago. It’s not that I’m completely against voice, but forcing it to be my *only* interface with a system is what I don’t like.
April 23rd, 2008 at 3:55 pm
Also, it would probably be possible to record you saying whatever it is the system is asking you to say at some point in your life. An unfortunate consequence of using audio biometrics is that everyone nearby can hear (and thus record) what you just spoke.