Invisible pixels protect South Korean graduates’ photos from deepfake abuse

Invisible pixels protect South Korean graduates’ photos from deepfake abuse

The alumni-built technology embeds photos with signals that confuse AI generators to prevent images from being manipulated

Seoul National University’s (SNU) use of StealCut Protect, created by a start-up founded by SNU alumni, marked the first time it was used for a university graduation album in the country, The Korea Herald reported. StealCut provided the technology to SNU free of charge.

Deepfake models generally work by analysing facial structure characteristics to generate manipulated content. StealCut Protect inserts minute, pixel-level signals – largely invisible to the human eye – into images to disrupt this process. If a deepfake generator tries to manipulate a protected photo, the result is a distorted or unidentifiable face.