Metraipolis › archive › dc_944262facf20
Tracing Provenance and Detecting Tampering with Complementary LLM Watermarks — arXiv:2608.12713
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.12713
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T04:18:04.388Z
- SHA-256 of the archived text
952fd923e2d900c83c6ca5d792c09d5b48c0e49b0daeb6222d9b8e8ad92ab541- Stable URL of this page
https://metraipolis.com/documento/dc_944262facf20
How to cite
Tracing Provenance and Detecting Tampering with Complementary LLM Watermarks — arXiv:2608.12713 — arXiv (en), https://arxiv.org/abs/2608.12713 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_944262facf20 (Metraipolis archive dc_944262facf20, sha256 952fd923e2d900c83c6ca5d792c09d5b48c0e49b0daeb6222d9b8e8ad92ab541)
BibTeX
@misc{dc944262facf20,
title = {Tracing Provenance and Detecting Tampering with Complementary LLM Watermarks — arXiv:2608.12713},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.12713},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_944262facf20; sha256 952fd923e2d900c83c6ca5d792c09d5b48c0e49b0daeb6222d9b8e8ad92ab541},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.12713. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Watermarking LLM-generated text is an important task for tracing its provenance. Existing LLM watermarks preserve provenance under editing, but this same robustness allows an adversary to alter critical content while retaining attribution, a vulnerability known as piggyback spoofing. We introduce an innovative watermark that jointly provides provenance and tamper evidence. It co-embeds a robust signal and a fragile signal into each generated token. The signals share the same mechanism but use independent keys and different seeding windows over normalized text, making one resilient to edits and the other sensitive to reader-visible changes. Multiple rounds of unbiased tournament reweighting preserve the expected generation distribution, while a periodic round-allocation pattern controls the trade-off between the two signals. At detection, their scores form a two-dimensional space supporting three decisions: Intact, Tampered, and No-Watermark. Across two large language models and two prompt datasets, our method demonstrates the highest tamper-detection rate among the evaluated methods while maintaining competitive attribution robustness and perplexity. Ablation studies show that reliable three-state detection requires a well-defined notion of intactness, co-embedding of the two signals, and complementary sensitivity to edits. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-14 in cs.AI). Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2608.12713 — 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 2026-10-19T04:13:06.690Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-14?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).