GIFT: Gradient-aware Immunization of diffusion models against malicious Fine-Tuning with safe concepts retention
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916849746182144 |
|---|---|
| author | Abdalla, Amro Shaheen, Ismail DeGenaro, Dan Mallick, Rupayan Raita, Bogdan Bargal, Sarah Adel |
| author_facet | Abdalla, Amro Shaheen, Ismail DeGenaro, Dan Mallick, Rupayan Raita, Bogdan Bargal, Sarah Adel |
| contents | We present GIFT: a {G}radient-aware {I}mmunization technique to defend diffusion models against malicious {F}ine-{T}uning while preserving their ability to generate safe content. Existing safety mechanisms like safety checkers are easily bypassed, and concept erasure methods fail under adversarial fine-tuning. GIFT addresses this by framing immunization as a bi-level optimization problem: the upper-level objective degrades the model's ability to represent harmful concepts using representation noising and maximization, while the lower-level objective preserves performance on safe data. GIFT achieves robust resistance to malicious fine-tuning while maintaining safe generative quality. Experimental results show that our method significantly impairs the model's ability to re-learn harmful concepts while maintaining performance on safe content, offering a promising direction for creating inherently safer generative models resistant to adversarial fine-tuning attacks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_13598 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | GIFT: Gradient-aware Immunization of diffusion models against malicious Fine-Tuning with safe concepts retention Abdalla, Amro Shaheen, Ismail DeGenaro, Dan Mallick, Rupayan Raita, Bogdan Bargal, Sarah Adel Cryptography and Security Artificial Intelligence Computer Vision and Pattern Recognition Machine Learning We present GIFT: a {G}radient-aware {I}mmunization technique to defend diffusion models against malicious {F}ine-{T}uning while preserving their ability to generate safe content. Existing safety mechanisms like safety checkers are easily bypassed, and concept erasure methods fail under adversarial fine-tuning. GIFT addresses this by framing immunization as a bi-level optimization problem: the upper-level objective degrades the model's ability to represent harmful concepts using representation noising and maximization, while the lower-level objective preserves performance on safe data. GIFT achieves robust resistance to malicious fine-tuning while maintaining safe generative quality. Experimental results show that our method significantly impairs the model's ability to re-learn harmful concepts while maintaining performance on safe content, offering a promising direction for creating inherently safer generative models resistant to adversarial fine-tuning attacks. |
| title | GIFT: Gradient-aware Immunization of diffusion models against malicious Fine-Tuning with safe concepts retention |
| topic | Cryptography and Security Artificial Intelligence Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2507.13598 |