Metraipolis › archive › dc_53367fa193a0

Region-Level Black-Box Defense Against Stealthy Embedding-Space Backdoors in CLIP — arXiv:2609.31558

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.31558
License (as archived)
CC0-1.0
Archived
2026-10-08T02:09:29.856Z
SHA-256 of the archived text
56cfac5620d9a5abb00c7e951c8c9f9a9dbdf298d0dea02404354846b2f5bd83
Stable URL of this page
https://metraipolis.com/documento/dc_53367fa193a0

How to cite

Region-Level Black-Box Defense Against Stealthy Embedding-Space Backdoors in CLIP — arXiv:2609.31558 — arXiv (en), https://arxiv.org/abs/2609.31558 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_53367fa193a0 (Metraipolis archive dc_53367fa193a0, sha256 56cfac5620d9a5abb00c7e951c8c9f9a9dbdf298d0dea02404354846b2f5bd83)
BibTeX
@misc{dc53367fa193a0,
  title = {Region-Level Black-Box Defense Against Stealthy Embedding-Space Backdoors in CLIP — arXiv:2609.31558},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.31558},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_53367fa193a0; sha256 56cfac5620d9a5abb00c7e951c8c9f9a9dbdf298d0dea02404354846b2f5bd83},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31558. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Contrastive Language--Image Pretraining (CLIP) has emerged as a dominant vision backbone due to its strong transferability and zero-shot capabilities. However, recent studies reveal a critical vulnerability: embedding-space backdoor attacks. By poisoning only a tiny fraction of image--text pairs, adversaries can implant stealthy triggers that induce targeted shifts in CLIP's joint embedding space. Unlike conventional backdoors that manipulate classifier logits, these attacks corrupt representations directly, making them highly effective under extremely low poisoning ratios and difficult to detect. Existing defenses require access to model parameters, gradients, logits, or clean validation data---assumptions that rarely hold in realistic black-box deployments. Moreover, current black-box methods struggle to accurately localize small or out-of-distribution triggers. We propose CLIPGuard, a lightweight and fully black-box defense specifically designed to mitigate embedding-space backdoors in CLIP encoders. CLIPGuard identifies malicious regions by measuring segment-wise embedding perturbations and selectively purifies only suspicious segments via semantic inpainting, preserving benign visual content and alignment quality. Extensive experiments on STL-10, ImageNet, and diverse trigger families---including BadCLIP, BadNets, blended, patch-based, and typographic attacks---demonstrate that CLIPGuard reduces attack success rates to as low as 1.05% while maintaining clean accuracy up to 86.34%, consistently outperforming existing black-box defenses, including CleanCLIP and CleanerCLIP. Our code is available this https URL

Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-28 in cs.CR).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2609.31558 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T01:42:25.581Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).