CSF: Black-box Fingerprinting via Compositional Semantics for Text-to-Image Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914484947255296 |
|---|---|
| author | Lee, Junhoo Koo, Mijin Kwak, Nojun |
| author_facet | Lee, Junhoo Koo, Mijin Kwak, Nojun |
| contents | Text-to-image models are commercially valuable assets often distributed under restrictive licenses, but such licenses are enforceable only when violations can be detected. Existing methods require pre-deployment watermarking or internal model access, which are unavailable in commercial API deployments. We present Compositional Semantic Fingerprinting (CSF), the first black-box method for attributing fine-tuned text-to-image models to protected lineages using only query access. CSF treats models as semantic category generators and probes them with compositional underspecified prompts that remain rare under fine-tuning. This gives IP owners an asymmetric advantage: new prompt compositions can be generated after deployment, while attackers must anticipate and suppress a much broader space of fingerprints. Across 6 model families (FLUX, Kandinsky, SD1.5/2.1/3.0/XL) and 13 fine-tuned variants, our Bayesian attribution framework enables controlled-risk lineage decisions, with all variants satisfying the dominance criterion. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_16363 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | CSF: Black-box Fingerprinting via Compositional Semantics for Text-to-Image Models Lee, Junhoo Koo, Mijin Kwak, Nojun Cryptography and Security Artificial Intelligence Computer Vision and Pattern Recognition Text-to-image models are commercially valuable assets often distributed under restrictive licenses, but such licenses are enforceable only when violations can be detected. Existing methods require pre-deployment watermarking or internal model access, which are unavailable in commercial API deployments. We present Compositional Semantic Fingerprinting (CSF), the first black-box method for attributing fine-tuned text-to-image models to protected lineages using only query access. CSF treats models as semantic category generators and probes them with compositional underspecified prompts that remain rare under fine-tuning. This gives IP owners an asymmetric advantage: new prompt compositions can be generated after deployment, while attackers must anticipate and suppress a much broader space of fingerprints. Across 6 model families (FLUX, Kandinsky, SD1.5/2.1/3.0/XL) and 13 fine-tuned variants, our Bayesian attribution framework enables controlled-risk lineage decisions, with all variants satisfying the dominance criterion. |
| title | CSF: Black-box Fingerprinting via Compositional Semantics for Text-to-Image Models |
| topic | Cryptography and Security Artificial Intelligence Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.16363 |