Making Sense of Blockchain for AI Deepfakes Technology
Abbas Yazdinejad, Reza M. Parizi, Gautam Srivastava, Ali Dehghantanha
Abstract
Deepfakes generally refers to a new breed of adversarial deep learning technology to create non-consensual contents (mostly videos) for nefarious purposes. Most researches focus on the `detection' of deepfakes using AI-assisted approaches to take on this problem. This has been the common method operandi used by researchers thus far. However, there is one missing aspect of the deepfake problem, which is `authentication'. Instead of attempting to detect what content is fake, in this paper, we focus on techniques to provide tamper-proof evidence of what content is real. Blockchain has been advocated to be helpful with the authentication aspect of many real-world scenarios. Despite the scattered efforts around such solutions, there are no studies that can shed light on where it makes sense to adopt blockchain technology to better take on the deepfake problem. This paper aims to provide a one-stop guide to using blockchain to navigate deepfake artificial intelligence. We discuss potential use cases and solutions to tackle deepfakes technology via blockchain functionalities and features.