Membership Inference Attacks for Face Images Against Fine-Tuned Latent Diffusion Models
This addresses privacy concerns for individuals whose face images are used in training datasets, but it is incremental as it applies existing attack methods to a specific model type.
This paper tackles the problem of privacy risks from generative image models by investigating membership inference attacks on face images used to fine-tune Latent Diffusion Models, finding that the proposed attack is viable in a realistic black-box setup with significant performance improvements from generated auxiliary data and watermarks.
The rise of generative image models leads to privacy concerns when it comes to the huge datasets used to train such models. This paper investigates the possibility of inferring if a set of face images was used for fine-tuning a Latent Diffusion Model (LDM). A Membership Inference Attack (MIA) method is presented for this task. Using generated auxiliary data for the training of the attack model leads to significantly better performance, and so does the use of watermarks. The guidance scale used for inference was found to have a significant influence. If a LDM is fine-tuned for long enough, the text prompt used for inference has no significant influence. The proposed MIA is found to be viable in a realistic black-box setup against LDMs fine-tuned on face-images.