Positive-Unlabeled Diffusion Models for Preventing Sensitive Data Generation
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916644584947712 |
|---|---|
| author | Takahashi, Hiroshi Iwata, Tomoharu Kumagai, Atsutoshi Yamanaka, Yuuki Yamashita, Tomoya |
| author_facet | Takahashi, Hiroshi Iwata, Tomoharu Kumagai, Atsutoshi Yamanaka, Yuuki Yamashita, Tomoya |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_03789 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Positive-Unlabeled Diffusion Models for Preventing Sensitive Data Generation Takahashi, Hiroshi Iwata, Tomoharu Kumagai, Atsutoshi Yamanaka, Yuuki Yamashita, Tomoya Machine Learning Artificial Intelligence 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. |
| title | Positive-Unlabeled Diffusion Models for Preventing Sensitive Data Generation |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2503.03789 |