New research from Florida State University and Rutgers University explores the potential of continuous liveness detection for voice biometrics on smart devices. The paper tries to establish a new level of security for voice biometrics as the technology is particularly vulnerable to spoofing through replay attacks, according to the researchers. 

Replay attacks try to spoof voice authentication systems by using a pre-recorded audio biometric sample.

The researchers designed the liveness detection tool, called VoiceGesture, to work on smart devices like smartphones and smart speakers. The solution aims to distinguish live users from recordings. The technology works by using the unique articulatory gesture of a user when they speak a passphrase, as well as audio hardware advances of smart devices.

According to the researchers, VoiceGesture does not require specialized hardware, but rather works via a speaker and a microphone available on smart devices supporting voice input.