Metraipolis › archive › dc_f2db2b23ae00

An Empirical Study of Output-to-Input Loops for Black-Box Backdoor Detection in Fine-Tuned Open-Weight LLMs — arXiv:2608.11348

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.11348
License (as archived)
CC0-1.0
Archived
2026-10-19T05:32:37.460Z
SHA-256 of the archived text
c774f8c07de5fe26cdc9daeb3ca6f79f472b34af74a40dd879f10acdcfb71488
Stable URL of this page
https://metraipolis.com/documento/dc_f2db2b23ae00

How to cite

An Empirical Study of Output-to-Input Loops for Black-Box Backdoor Detection in Fine-Tuned Open-Weight LLMs — arXiv:2608.11348 — arXiv (en), https://arxiv.org/abs/2608.11348 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f2db2b23ae00 (Metraipolis archive dc_f2db2b23ae00, sha256 c774f8c07de5fe26cdc9daeb3ca6f79f472b34af74a40dd879f10acdcfb71488)
BibTeX
@misc{dcf2db2b23ae00,
  title = {An Empirical Study of Output-to-Input Loops for Black-Box Backdoor Detection in Fine-Tuned Open-Weight LLMs — arXiv:2608.11348},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.11348},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f2db2b23ae00; sha256 c774f8c07de5fe26cdc9daeb3ca6f79f472b34af74a40dd879f10acdcfb71488},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.11348. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Anyone can upload a fine-tuned large language model (LLM) to a public repository and claim it is safe. A backdoored model behaves normally on ordinary inputs until a hidden trigger fires, and a user with no training data, clean reference weights, or the trigger phrase has no clear way to check the model before using it. We introduce and empirically evaluate self-feeding, a black-box test method that feeds a model's own output back as its next input, so the text drifts away from the starting prompt and toward the data the model was fine-tuned on. We test self-feeding against a repeated same-prompt baseline on six open-weight LLMs (3B-15B parameters), each fine-tuned with backdoors spanning eleven attack categories, using twenty ordinary starting prompts and chains of up to ten steps. Self-feeding finds backdoors in five of six models at 92.0\% pooled precision, while the same-prompt baseline succeeds on only one of 120 prompt-model pairs; chains that begin with a joke request, an arithmetic question, or a coffee recipe all reach a trigger within a few steps. Recall per prompt is low (19.2\%), and we show why it still adds up to much higher detection at the model level once several starting prompts are used. We also report where the method falls short: one model was never triggered, and self-feeding produced two false positives that the same-prompt baseline cannot produce. Cutting the chains to four steps keeps every model-level detection at 100\% precision while using 60\% fewer queries. Needing only text-level query access and a way to recognize malicious output, self-feeding offers a cheap first check on a downloaded model.

Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-08-13 in cs.CR).
Categorie arXiv: Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2608.11348 — 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 2026-10-19T05:17:58.116Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-08-13?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).