Panel: PAN-4: Beyond Words: Recognition, Spoofing, and Anonymization of Individual Traits in Speech

Speech is among the most natural and convenient means of biometric authentication. The individual traits embedded in the speech signals form the basis of speaker recognition or voice authentication. With the widespread availability of speech synthesis tools, the threat from spoofing attacks to speaker recognition systems is growing since fraudsters can use these tools to produce a natural-sounding speech of a victim. While research on speech anti-spoofing has seen significant progress in the past few years, privacy concerns have called for the need for speech anonymization. In this panel, we invite world-leading experts to share their opinions on the security and the privacy expects in handling individual traits in speech, the challenges posed by the advancement in neural speech synthesizers, and the collaborative efforts that could be put together in answering the concerns and challenges. Format 10 minutes of introduction (moderator) 30 minutes of presentation (panelists) 50 minutes of open discussion Call for contribution to ASVspoof5