Deepfake Style Transfer Mixture: a First Forensic Ballistics Study on Synthetic Images
International Conference on Image Analysis and Processing (ICIAP), 2022
Luca Guarnera1,2, Oliver Giudice1,3, Sebastiano Battiato1,2
1 Department of Mathematics and Computer Science, University of Catania, Italy
2 iCTLab s.r.l. Spinoff of University of Catania, Italy
3 Applied Research Team, IT dept., Banca d'Italia, Italy
luca.guarnera@unict.it, {giudice, battiato}@dmi.unict.it
Style transfer operations. H(.) represents the function that computes the RGB histogram.
ABSTRACT
Most recent style-transfer techniques based on generative architectures are able to obtain synthetic multimedia contents, or commonly called deepfakes, with almost no artifacts. Researchers already demonstrated that synthetic images contain patterns that can determine not only if it is a deepfake but also the generative architecture employed to create the image data itself. These traces can be exploited to study problems that have never been addressed in the context of deepfakes. To this aim, in this paper a first approach to investigate the image ballistics on deepfake images subject to style-transfer manipulations is proposed. Specifically, this paper describes a study on detecting how many times a digital image has been processed by a generative architecture for style transfer. Moreover, in order to address and study accurately forensic ballistics on deepfake images, some mathematical properties of style-transfer operations were investigated.
|
Download Paper |
Datasets Details
Different Deepfake images were generated considering StarGAN-v2 architecture:
Dataset used to define whether a multimedia content has undergone one or two deepfake manipulation operations belonging to the style-transfer category
Dataset used to demonstrate several mathematical properties of the style transfer operation