Please Say Your Password At The Beep

Article written by guest writer Kecia Lynn

What’s the Latest Development?

Next week in Germany a team of computer scientists from Carnegie Mellon University will present a paper describing a new method of voice data encryption that involves matching it against a record that does not contain an actual sample of the the speaker’s voice. Their approach turns the user’s voice sample (which is taken when he or she first registers at a location where identification is required) into a kind of password itself, and creates multiple unique chunks of sample data that can be used to identify the user when he or she says an arbitrary phrase.

What’s the Big Idea?

Biometric approaches to security have been supported by researchers and others for years, but the level to which such data can be made truly secure has long posed a problem. Voice data in particular can reveal too much information, which could then be easily matched against preexisting recordings on less-secure sites such as YouTube. The team also took into consideration the amount of time spent decoding thousands of simultaneous logins on the server side, by creating a separate piece of code from all those previously-generated chunks of sample data and using it as the basis for comparison. 

Photo Credit: Shutterstock.com