ADOBE has developed a tool that can detect if an image has been tampered with.
The AI could tell if an element had been added, moved or cut from a photo.
But the company warned that no piece of technology could provide a foolproof verification system.
Vlad Morarium, an Adobe researcher, employed artificial intelligence to scan for signs of manipulation that are not usually visible to the naked eye.
Photoshop, created 28 years ago, is a powerful image editor, and its name has become a verb for image manipulation.
Existing verification tools can scan an image file's metadata - which contains information on when and where a photo was taken - for signs of mischief, and look for things like inconsistent lighting.
Mr Morarium, who spent 14 years researching ways to spot image manipulation, taught an artificial intelligence network to recognise signs of colour change and noise inconsistencies in tens of thousands of pictures.
The initial study focused on three common manipulation techniques:
- splicing, where parts of two different images are combined
- removal, where an object is removed from a photograph, and filled in
- copy-move, where objects in a photograph are moved or cloned from one place to another
"Each of these techniques tend to leave certain artefacts, such as strong contrast edges, deliberately smoothed areas, or different noise patterns," he notes.
Mr Morarium, whose research was carried out in conjunction with the US government agency Darpa, said the algorithm might also detect differences in illumination and unusual compression in the future.
He added that Adobe, which brought image manipulation capabilities to the masses, was "uniquely positioned" to create tools to determine authenticity.
No comments:
Post a Comment