Deepfake Detection: Methods for Combating and Detecting Deepfakes
Elizabeth Galoozis, Associate University Librarian and Head, Information Literacy
Curtis Fletcher, Director, Ahmanson Lab
Samir Ghosh, Project Coordinator, Ahmanson Lab
The term deepfake, a combination of “deep learning” and “fake,” designates a new class of hyper-realistic, fake media - images, video, and audio generated leveraging machine-learning algorithms to, primarily, superimpose faces and/or voices on people in order to manufacture their saying and doing things they have not said or done.
While there are real artistic benefits to such techniques, much has recently been made about the great social harm the technology can unleash in an increasingly polarized and fragmented social media ecosystem.
In this workshop, we’ll discuss with participants the troubling nature of truth in an era of deepfakes. We’ll survey the history of visual evidence and media manipulation; introduce participants to current large-scale media forensic efforts to combat deepfakes; talk about the methods, techniques, and technology behind the creation of deepfakes; and, finally, offer some best practices for researching and spotting AI-generated fake media online.