My name is Brandon Parker. I am a PhD student working on a dissertation about the possibilities of using deepfakes in phishing attacks. I, myself have never worked with deepfake tooling, but understand machine learning and the concepts behind it.
Because I have not actually tried to create a deepfake, I am unsure of how much work is involved in producing a believable deepfake (video or audio). In other words, I am exploring to what lengths an attacker would need to go to utilize this technology in this way for it to be effective.
For my study, I am thinking about creating both video and audio deepfakes of a person(s) and then real audio/video of the real person and let subjects try and determine which is real. Studies of this nature have done before like this.
Any feedback you would like to provide on this topic would be appreciated!