Facebook: In partnership with Michigan State University (MSU), in the United States, Facebook researchers are developing a new tool capable of identifying deepfakes and even determining their origin, facilitating the removal of these contents. The novelty was presented by the social network this Wednesday (16).
Deepfake: what is it and how does it work?
Increasingly evolved, deepfake creation technology uses artificial intelligence (AI) to make very realistic montages. They can be used to deceive large numbers of people, spreading fake news across different platforms and attacking the reputation of well-known figures.
Using reverse engineering, the software created by scientists at MSU and Facebook analyzes the patterns of the AI that generated the assembly. “We can estimate properties of the generative models used to create each deepfake and even associate multiple deepfakes with the model that possibly produced them,” explained research co-author Tal Hassner.
In this way, the system can identify some imperfections that arise during the editing of the content, modifying the “digital print” of the image. From this change, the algorithm is able to differentiate fake photos and videos from real ones and suggest which neural network was used in the creation of the deepfake.