SEI Cyber Talk (Episode 13). Deepfakes: What Can Really Be Done Today?
Abstract
Rotem Guttman: Hi, this is Rotem Guttman, and I'm going to be speaking with a fellow researcher, Zach Kurtz, about deep fakes, what's real and what's fake. So Zach, to jump right in, what are deep fakes? Zach Kurtz: All right. Well, a deep fake is any kind of-- typically photos, but it really could be any kind of essays or any media-- that has--Rotem Guttman: Or videos. Zach Kurtz: Videos or photos, right, that has some fake content in it, and it's not just any fake content. There's all kinds of ways to just throw some weird pixels or content into a bit of media, but a deep fake uses a deep neural network which is a kind of machine learning thing where they can generate fake content and it's really taken off in the last few years. And so the other aspect of it is it's not just how you make it, but how you use it. So you could imagine using some of this modern machine learning stuff to do red eye removal or do other special effects on images that serve a simple artistic purpose or help to make the image more truthful to the original goal, but with a deep fake, it's being used to trick or deceive or even to humor people, and that's part of the important part of what makes up a deep fake.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2021
- Accession Number
- AD1128168
Entities
People
- Rotem Guttman
- Zach Kurtz
Organizations
- Carnegie Mellon University