Hierarchical Anti-Aesthetics: Protecting Facial Privacy against Customized Diffusion Models

Researchers have introduced a method called Hierarchical Anti-Aesthetics to protect personal photos from being used to train customized generative AI models. By applying subtle, imperceptible modifications to images, the technique disrupts the ability of diffusion models to learn and reconstruct specific faces. This approach aims to provide individuals with more control over their biometric privacy as personalized image generation technology becomes more accessible.