Deepfake picture editing makes it possible to create evidence of scenes that never happened. Concerns about deepfakes have led to a number of countermeasures, the latest being Markpainting.
Markpainting is a technique that watermarks photos, this makes it easy to detect ML-based manipulations.
Users can modify images in a number of ways which makes it difficult to detect. While the mark painter sabotages any attempt to paint it as it adds visible information.
It uses tamper-resistant marks which ensure watermarks are more robust. For example, a photo agency that makes stock images with copyright watermarks for its website can markpaint them. This will prevent anyone from removing the watermark using common editing software.
You can learn more about markpaint in the paper “Markpainting: Adversarial Machine Learning Meets Inpainting” by David Khachaturov, Ilia Shumailov, Yiren Zhao, Nicolas Papernot, and Ross Anderson.