Fraunhofer IOSB RealorRender tool aims to flag AI deepfakes with hybrid reconstruction

Researchers at Germany's Fraunhofer Institute of Optronics, System Technologies, and Image Exploitation (IOSB), with funding from the German Federal Office for Information Security (BSI), have developed RealorRender, a system that classifies images as authentic or AI-generated and explains the decision, New Atlas reported. Senior scientist Andreas Specker said the method first uses an AI image generator to reconstruct the image, then applies a hybrid calculation of reconstruction error so the system can report recognition accuracy as a percentage. When a generator successfully clones a source image, that success is treated as evidence the source was a deepfake. New Atlas said RealorRender's success rate starts at 85% and peaks at 91%, with Specker noting it can be higher in individual cases. The write-up places the work against deepfake-driven harassment and fraud, citing Gov.UK figures on sexual deepfakes and UK ministerial pledges on nude images and phones, while arguing detection tools remain necessary beyond device rules.