While our voices are increasingly being used for biometric identification, they're not an infallible form of ID. People can impersonate us, for instance, or edited recordings of our voices can be used. It was with this in mind that researchers at the University of Michigan created VAuth ("vee-auth"), which adds an extra level of security to the technology.
Developed by a team led by Prof. Kang Shin, the hardware end of VAuth can take the form of a necklace, ear buds or an attachment to eyeglasses – all three have been buit. Whatever the form, it incorporates an accelerometer that detects subtle vibrations in the skin of the wearer's face, throat or chest as they speak.
The way in which precise vibrations match up to specific sounds make up a signature that's unique to each person. The idea is that VAuth would be combined with conventional voice-print identification, to make an extra-secure hybrid system.
In lab tests utilizing 18 users and 30 voice commands, VAuth managed a detection accuracy rate of 97 percent – this was regardless of where it was worn on the body, along with the user's language, accent or mobility. It also stood up to various practical attacks, such as impersonations or playback of voice recordings.
"Increasingly, voice is being used as a security feature but it actually has huge holes in it," says Shin. "If a system is using only your voice signature, it can be very dangerous. We believe you have to have a second channel to authenticate the owner of the voice."
The university has applied for a patent, and is currently seeking commercial partners to help bring VAuth to market.