Researchers explore continuous liveness detection for voice biometrics
New research from Florida State University and Rutgers University scientists has explored the potential of continuous liveness detection for voice biometrics on smart devices.
Replay attacks involve a malicious adversary trying to spoof voice authentication systems by using a pre-recorded voice biometric sample collected from a genuine user.
Dubbed VoiceGesture, the liveness detection tool designed by the researchers is intended to work on smart devices like smartphones and smart speakers.
The solution is capable of distinguishing a live user from a recording by utilizing both the unique articulatory gesture of the user when they speak a passphrase, as well as the audio hardware advances of smart devices.
From a technical standpoint, VoiceGesture repurposes the pair of built-in speakers and microphone on a smart device as a Doppler radar, which in turn transmits a high-frequency acoustic sound from the speaker while also listening to the reflections at the microphone when a user speaks a passphrase.
The Doppler shifts resulting from the user’s articulatory gestures are then extracted and used for biometric liveness detection.
According to the researchers, VoiceGesture represents a practical solution as it does not require any specialized hardware, working instead via a speaker and a microphone commonly available on smart devices supporting voice input.
“Our experimental evaluation with 21 participants and different smart devices shows that VoiceGesture achieves over 99% and around 98% detection accuracy for text-dependent and text-independent liveness detection, respectively,” the researchers said in the introduction to the paper.
Additional results showed that VoiceGesture is resilient to different device placements, as well as low audio sampling frequency.
The platform also supports medium-range liveness detection on smart speakers in various use scenarios, including smart homes and smart vehicles.