PhotoGuard AI

PhotoGuard AI is a research project from MadryLab that aims to protect images from malicious AI-powered editing. It is designed for individuals and companies who want to safeguard their images from being manipulated by generative models. The project focuses on latent diffusion models, such as Stable Diffusion, and uses adversarial attacks to make images immune to direct editing.
The core of PhotoGuard AI's 'immunization' process is to leverage so-called adversarial attacks on these generative models. By slightly modifying the input image, PhotoGuard AI can create a 'guard' that makes the image immune to direct editing by generative models. This is achieved by modifying the starting image with the goal of breaking the whole end-to-end diffusion process.
PhotoGuard AI is particularly useful for individuals and companies who want to protect their images from being manipulated for malicious purposes. For instance, news organizations can use PhotoGuard AI to safeguard their images from being edited to spread misinformation. Similarly, individuals can use PhotoGuard AI to protect their personal images from being manipulated and shared without their consent.