Breaking voice authentication - Security testing approach

Abstract

A couple of years ago, along with raising popularity of smart devices that promote voice authentication mechanisms, voice recognition became a buzz topic again. Not only voice patterns recognition algorithms significantly improved their quality, but also with a rapid growth in the development of sound editing tools and an outburst of the deep-fake concepts, the attack surface of voice authenticators significantly expanded. This paper describes the approach for penetration testing of the voice recognition solutions and tackles common misconceptions of voice patterns characteristics.