Post · bonfire.cafe

Fabrizio Musacchio

@FabMusacchio@mastodon.social · 16 hours ago

#Deepfakes are everywhere, but #DigitalForensics investigators are fighting back:

🌍 https://scim.ag/42dMPBg

3 media

To verify images, digital forensics investigators will often check whether the geometry of the scene is realistic. In a real photo, lines that run parallel in reality—like floor tiles—should meet at a single vanishing point. In this image, however, the dotted lines do not meet in a single point, indicating it is a fake.
Image: an AI-generated image of soldiers marching down a hallway in three lines, wearing fatigues and carrying rifles. Four dotted white lines have been added over the image, running along the floor tiles visible in the foreground and extending back to where the vanishing point should be, behind the soldiers. Instead of meeting in a single point, the lines all cross at different points. — To verify images, digital forensics investigators will often check whether the geometry of the scene is realistic. In a real photo, lines that run parallel in reality—like floor tiles—should meet at a single vanishing point. In this image, however, the dotted lines do not meet in a single point, indicating it is a fake. Image: an AI-generated image of soldiers marching down a hallway in three lines, wearing fatigues and carrying rifles. Four dotted white lines have been added over the image, running along the floor tiles visible in the foreground and extending back to where the vanishing point should be, behind the soldiers. Instead of meeting in a single point, the lines all cross at different points.

Investigators also examine reflections. The lines connecting points on an object to matching points in its mirror image run parallel in reality, so similarly should meet at a vanishing point. In this image, the lines again do not converge on one point, revealing it to be a fake.
Image: An AI-generated image of a plastic cartoon dinosaur toy with its reflection visible in a small mirror. Four dotted white lines are layered over the image, connecting points on the toy with the same points on its reflection: the top of its eye, the end of its jaw, its hand, and its foot. The lines extend out to the side of the image, where they all cross at different points rather than meeting at a single point. — Investigators also examine reflections. The lines connecting points on an object to matching points in its mirror image run parallel in reality, so similarly should meet at a vanishing point. In this image, the lines again do not converge on one point, revealing it to be a fake. Image: An AI-generated image of a plastic cartoon dinosaur toy with its reflection visible in a small mirror. Four dotted white lines are layered over the image, connecting points on the toy with the same points on its reflection: the top of its eye, the end of its jaw, its hand, and its foot. The lines extend out to the side of the image, where they all cross at different points rather than meeting at a single point.

#digitalforensics #deepfakes

truh

@truh@shark.community · 1 minute ago

@FabMusacchio especially with the first image the building might be a bit crooked or the lines might just be a bit off. It's pretty close to having a vanishing point. It's even closer to having a vanishing point if you remove the wall line, that might not be perfectly parallel and look just at the tiles.

lps

@lps@mograph.social · 54 minutes ago

@FabMusacchio thanks for reposting, these are great tips to help spot fakes:) it's getting incredibly hard, if not impossible lately: (

Ragon2

@Ragon2@mstdn.social · 3 hours ago

@FabMusacchio Great to have a sure way to prove things, but honestly, just zoom in on details see if they're coherent? In most if not any genai image? Not even talking about the dinausore/crocodile one because if you're under 50 you could tell in one blink

2 media

jose

@castrillo@mastodon.online · 4 hours ago

@FabMusacchio good post 👍

Doesn't Exist.

@Deixis9@beige.party · 4 hours ago

@FabMusacchio @spellingmistakescostlives

·J Mopp

@jmopp@masto.ai · 4 hours ago

@FabMusacchio

An image macro of a guy pointing at a computer screen. The caption says: This looked slopped. I can tell from the lines of perspective and from seeing quite a few slops in my time

Ilusënn

@ilusenn@corneill.es · 5 hours ago

@FabMusacchio Hello! Can I ask where the url shortener sends?

Xauri'EL Zwaan

@XauriEL@mastodon.nz · 5 hours ago

@FabMusacchio it is really telling that one of the most powerful tools to detect fake images is an artistic tool that any first year art school student knows about

F4GRX SÃ©bastien

@f4grx@chaos.social · 5 hours ago

@FabMusacchio beautiful analysis.

Microblog Castellano

@microblogc@neopaquita.es · 6 hours ago

@FabMusacchio So if I want to commit a murder, I have years to prepare it and I know the place will be surveilled with cameras, I should pave it with slightly non-parallell tiles, to get a plausible deniability.

ChookMother 🇦🇺🦘

@anne_twain@theblower.au · 6 hours ago

@FabMusacchio Interesting. Thanks for posting!

Mans R

@mansr@society.oftrolls.com · 7 hours ago

@FabMusacchio How does this method handle lens distortion?

pulsation

@pulsation@mamot.fr · 7 hours ago

@FabMusacchio
Very interesting, thanks.
@defakator it might be of your interest?

Aude Caussarieu 🌻

@AudeCaussarieu@social.sciences.re · 8 hours ago

Ah bah voilà une super source d'exos d'optique géométrique 😅

Dr Grototo

@grototo@social.sciences.re · 7 hours ago

@AudeCaussarieu Je me demande quand même à quel point c’est difficile de distinguer de telles anomalies de distorsions optiques causées par la lentille, qui va courber les lignes parallèles.

Chèvre Mousse

@legendarybassoon@mamot.fr · 7 hours ago

@grototo @AudeCaussarieu On verrait un autre genre de déformation, pas des trucs au pif comme ça, je dirais ? De toute façon dans le genre de photo montré est rarement prise en grand angle...

Nartagnan ⏚

@nartagnan@mstdn.fr · 5 hours ago

@legendarybassoon @grototo @AudeCaussarieu

Oui, je suppose ça ferait une forme pas exacte, mais pas random.
L'images des cubes par exemple, il y aurait de la symétrie, pas juste un des points qui va n'importe où.

Et j'imagine la gueule des prompts pour dire de se concentrer sur les points de fuite, difficile à décrire en langage "naturel". Avec un peu de chances ça va pas être facile à corriger.

2 more replies

Mathieu

@mathieu_@eldritch.cafe · 8 hours ago

@AudeCaussarieu
J'aurais jamais pensé que les points de fuites pouvaient être une façon de vérifier la véracité d'une image. 🤯