Perception-guided Jailbreak against Text-to-Image Models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909484703547392 |
|---|---|
| author | Huang, Yihao Liang, Le Li, Tianlin Jia, Xiaojun Wang, Run Miao, Weikai Pu, Geguang Liu, Yang |
| author_facet | Huang, Yihao Liang, Le Li, Tianlin Jia, Xiaojun Wang, Run Miao, Weikai Pu, Geguang Liu, Yang |
| contents | In recent years, Text-to-Image (T2I) models have garnered significant attention due to their remarkable advancements. However, security concerns have emerged due to their potential to generate inappropriate or Not-Safe-For-Work (NSFW) images. In this paper, inspired by the observation that texts with different semantics can lead to similar human perceptions, we propose an LLM-driven perception-guided jailbreak method, termed PGJ. It is a black-box jailbreak method that requires no specific T2I model (model-free) and generates highly natural attack prompts. Specifically, we propose identifying a safe phrase that is similar in human perception yet inconsistent in text semantics with the target unsafe word and using it as a substitution. The experiments conducted on six open-source models and commercial online services with thousands of prompts have verified the effectiveness of PGJ. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_10848 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Perception-guided Jailbreak against Text-to-Image Models Huang, Yihao Liang, Le Li, Tianlin Jia, Xiaojun Wang, Run Miao, Weikai Pu, Geguang Liu, Yang Computer Vision and Pattern Recognition In recent years, Text-to-Image (T2I) models have garnered significant attention due to their remarkable advancements. However, security concerns have emerged due to their potential to generate inappropriate or Not-Safe-For-Work (NSFW) images. In this paper, inspired by the observation that texts with different semantics can lead to similar human perceptions, we propose an LLM-driven perception-guided jailbreak method, termed PGJ. It is a black-box jailbreak method that requires no specific T2I model (model-free) and generates highly natural attack prompts. Specifically, we propose identifying a safe phrase that is similar in human perception yet inconsistent in text semantics with the target unsafe word and using it as a substitution. The experiments conducted on six open-source models and commercial online services with thousands of prompts have verified the effectiveness of PGJ. |
| title | Perception-guided Jailbreak against Text-to-Image Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2408.10848 |