98 views
Today, with Artificial Intelligence, you can create voice messages by cloning anyone's voice. Artificial audio is increasingly precise, and can even simulate unique speaker characteristics, such as inflections, in addition to timbre. An audio authentication expertise in the field of deep fake audio can be particularly complex, since the vocal characteristics will over time be almost identical between an artificially generated message and a genuine one. For this reason, it was decided to have the machine, the software, do the work, which does not "listen" to the audio but analyses it numerically by comparing the values of various features of multiple genuine vocal samples, with the values of the features of the suspect audio. In the talk, Doctor Bassetti talks about how he created the software and what it does, and forensic audio expert Marco Perino explains the development potential of an analysis system of this type, and the current critical issues with traditional tools.