Adobe’s senior research scientist, Vlad Morariu, employed AI to scan a couple of images to find out manipulations in the images that the naked out could not notice, according to a blog post by Adobe.
Ever since Adobe Photoshop came into our lives, people have edited images for artistic expression, but more popularly to manipulate images for deception, becoming a synonym for edited images.
However, the company decided to take it upon itself to fix the deception problem and now AI can tell you if an element has been added, moved or cut out from a photograph. They also warned that no piece of technology can provide a foolproof verification system.
“File formats contain metadata that can be used to store information about how the image was captured and manipulated. Forensic tools can be used to detect manipulation by examining the noise distribution, strong edges, lighting and other pixel values of a photo. Watermarks can be used to establish the original creation of an image,” said Marariu explaining that a variety of tools already were in existence to help document and trace the digital manipulation of photos.
However, none of these tools completely help detect enhancements in photographs. AI and machine learning, on the other hand, detect image enhancements faster.
Three common techniques were looked at. The first one was splicing, where two different images are combined. The second one copy-move, where objects in one photograph are moved or cloned. And the third was removal, where an object is removed, and space is filled in.
All these techniques leave behind some clues, and AI can successfully identify them using two methods, the first one using RGB stream and the second using noise stream filter.
Morris later explained that the software which brought image manipulation capabilities to the masses was "uniquely positioned" to create tools to determine authenticity.
In the future, Adobe aspires to extend this technology to be able to detect other artifacts of manipulation.