Metraipolis › archive › dc_85a51258620d

Clean Data Can Still Carry Backdoors: Support-Persistent Backdoors for Model Reuse — arXiv:2610.21208

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.21208
License (as archived)
CC0-1.0
Archived
2026-10-16T10:19:17.915Z
SHA-256 of the archived text
73dede77c06007f5a2ef4d5087842ec2a99fe8c36da51c1411ac9ef8a6fae6e6
Stable URL of this page
https://metraipolis.com/documento/dc_85a51258620d

How to cite

Clean Data Can Still Carry Backdoors: Support-Persistent Backdoors for Model Reuse — arXiv:2610.21208 — arXiv (en), https://arxiv.org/abs/2610.21208 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_85a51258620d (Metraipolis archive dc_85a51258620d, sha256 73dede77c06007f5a2ef4d5087842ec2a99fe8c36da51c1411ac9ef8a6fae6e6)
BibTeX
@misc{dc85a51258620d,
  title = {Clean Data Can Still Carry Backdoors: Support-Persistent Backdoors for Model Reuse — arXiv:2610.21208},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21208},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_85a51258620d; sha256 73dede77c06007f5a2ef4d5087842ec2a99fe8c36da51c1411ac9ef8a6fae6e6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21208. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Traditional backdoor attacks inject artificial trigger patterns into training data, causing models to misclassify triggered inputs while maintaining normal behavior on clean samples. Since these artificial triggers are absent in clean datasets, standard knowledge distillation (KD) typically eliminates them, leading to the common belief that KD serves as an effective purification defense. In this paper, we propose DAR (Discover-and-Relabel), a support-persistent backdoor method that retains its behavior even after clean-data distillation. DAR identifies naturally abundant patterns in low-dimensional feature subspaces using subspace clustering and applies label-only poisoning to samples that inherently satisfy the discovered rule. At test time, the backdoor is activated by locally modifying the corresponding subspace of the target images. By instantiating spatial and frequency operators, DAR achieves attack accuracy comparable to state-of-the-art backdoor attacks in ImageNet classification and CLIP-based prompt tuning. Notably, DAR keeps attack success above 96% after clean KD and remains defense-resistant in practice. Project page: this https URL

Tema degli esploratori: fiducia digitale e crittografia (elenco arXiv cs.CR, nuovi lavori del giorno).
Categorie arXiv: Cryptography and Security (cs.CR) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.21208 — 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 2026-10-16T08:47:32.599Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).