Voice Biometrics and Emerging Security Threats in the Voice Channel
MetadataShow full item record
Voice Biometrics is the automated recognition of individuals based on their voice characteristics that are both biological and behavioral. Voice Biometrics is currently deployed in many real world applications for personalization like voice assistants (e.g. Alexa and Google Home) or authentication and fraud prevention (e.g. in call centers). During this talk, we're going to describe how a voice biometrics system works, what are its main different components? How deep learning is helping this task? What are the current research challenges? In covid-19 era, how does wearing a mask affect the accuracy of voice biometrics? Additionally, we will focus on the emerging threats in the voice channel including deepfakes and replay attacks. We will finally describe our attempts to secure voice biometrics against those vulnerabilities.