In this article Dr. Bhusan Chettri provides an overview of how voice authentication systems can be compromised through spoofing attacks. He adds "spoofing attack refers to the process of making an un-authorised attempt to break into someone else's authentication system either using synthetic voices produces through AI technology or by performing a mimicry or by simply replaying a pre-recorded voice samples of the target user."
In this presentation, Bhusan Chettri provides an overview of voice authentication system that is based on automatic speaker verification technology. He provides background on both the traditional approaches of modelling speakers and current deep learning based approaches. A brief introduction to how these systems can be manipulated is also provided.
n this article, Dr. Bhusan Chettri provides a summary of ASVspoof - automatic speaker verification spoofing and countermeasures challenge - which is a biannual competition organised by the Speaker verification community to promote research and awareness in spoofing attacks towards voice authentication system while encouraging participation in the challenge to advance the research field in this topic. As part of the challenge, big-data to train and tune machine learning models for spoofing detection is released free of cost !!