Professionals all over the world are becoming more worried about the new AI tools that are making it a lot more easier to edit images and videos. Some of these tools happen to be developed by Adobe, still Adobe seems to be experimenting on a research about how machine learning could be used to identify edited pictures.
This the company demonstrated at the CVPR computer vision conference and showing the way machines could mechanically do digital forensics in much less time than is done by humans.
A spokesperson of the company stated that this venture was an “early-stage research project,” but the company intends that in the future it would be able to play a role in “developing technology that helps monitor and verify authenticity of digital media.”
The new research paper reveals how machine learning can be used to identify three major forms of image manipulation one of which is splicing – two parts of varying images combined, another, cloning – copying and pasting of objects within an image, and finally removal – an overall editing of an image.
To identify an edited image, professionals in digital forensics normally search for anything that may give away the fact that an image is edited in the hidden layers of the image. When an image is edited, a digital artifact is left behind, say inconsistencies in the random variations in the colour and brightness produced by image sensors also called image noise.
In a comment, digital forensics specialist, Hany Farid stated that, “The benefit of these new ML approaches is that they hold the potential to discover artifacts that are not obvious and not previously known. The drawback of these approaches is that they are only as good as the training data fed into the networks, and are, for now at least, less likely to learn higher-level artifacts like inconsistencies in the geometry of shadows and reflections.”
Setting aside the stipulations, it is kind of interesting to see that research is being conducted to aid in the spotting of digital fakes. With the way the world is going, there is need for a tool such as this to enable one differentiate between real and fake.