Positive-Unlabeled Diffusion Models for Preventing Sensitive Data Generation
This addresses a privacy and safety issue for users of generative AI by preventing unwanted sensitive data generation, though it is an incremental improvement as it builds on existing diffusion model frameworks.
The paper tackles the problem of diffusion models generating unwanted sensitive data by proposing positive-unlabeled diffusion models that use a small amount of labeled sensitive data and unlabeled data to prevent such generation, demonstrating through experiments that it can block sensitive images without reducing image quality.
Diffusion models are powerful generative models but often generate sensitive data that are unwanted by users, mainly because the unlabeled training data frequently contain such sensitive data. Since labeling all sensitive data in the large-scale unlabeled training data is impractical, we address this problem by using a small amount of labeled sensitive data. In this paper, we propose positive-unlabeled diffusion models, which prevent the generation of sensitive data using unlabeled and sensitive data. Our approach can approximate the evidence lower bound (ELBO) for normal (negative) data using only unlabeled and sensitive (positive) data. Therefore, even without labeled normal data, we can maximize the ELBO for normal data and minimize it for labeled sensitive data, ensuring the generation of only normal data. Through experiments across various datasets and settings, we demonstrated that our approach can prevent the generation of sensitive images without compromising image quality.