Metraipolis › archive › dc_f9780fe56edc

The Poisoned Conversation: Privacy-Leaking Watermarks in Unified Multimodal Models — arXiv:2610.05453

Source
arXiv
Original
https://arxiv.org/abs/2610.05453
License (as archived)
CC0-1.0
Archived
2026-10-06T05:29:52.458Z
SHA-256 of the archived text
6d87e08e8bee221c7d4519c075e1e812ad89a8ae208cbb89496d366d0c1cba30
Stable URL of this page
https://metraipolis.com/documento/dc_f9780fe56edc

How to cite

The Poisoned Conversation: Privacy-Leaking Watermarks in Unified Multimodal Models — arXiv:2610.05453 — arXiv, https://arxiv.org/abs/2610.05453 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f9780fe56edc (Metraipolis archive dc_f9780fe56edc, sha256 6d87e08e8bee221c7d4519c075e1e812ad89a8ae208cbb89496d366d0c1cba30)
BibTeX
@misc{dcf9780fe56edc,
  title = {The Poisoned Conversation: Privacy-Leaking Watermarks in Unified Multimodal Models — arXiv:2610.05453},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.05453},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f9780fe56edc; sha256 6d87e08e8bee221c7d4519c075e1e812ad89a8ae208cbb89496d366d0c1cba30},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.05453. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Multimodal models are increasingly shifting toward unified architectures that understand and generate text, images, and other modalities within a shared conversational context. This design enables fluid interaction across modalities, but it also changes the privacy threat model: Information revealed in one part of a conversation may remain accessible when the model later generates content in another modality. This risk is particularly concerning in settings where users rely on locally deployed models for privacy, assuming that sensitive interactions remain confined to their device. We introduce Privacy-Leaking Watermarks (PLWs): invisible, trigger-dependent watermarks that a malicious model provider can condition on prior chat history. With this adversarial intervention, the usual separation breaks: a sensitive keyword or semantic cue mentioned earlier in the conversation can cause a later, unrelated image to carry a hidden yet detectable watermark. PLWs pose a novel threat to users of unified multimodal models: A poisoned model can retain utility while covertly turning image generation into a channel for privacy leakage, even when deployed locally. Across 13 sensitive-attribute triggers and two model families, PLWs reach up to 100.0% TPR at 1% FPR. For example, across all tested conversational separations, OmniGen2 detects every prior disclosure of depression while falsely flagging only 1% of images generated without such a disclosure.

Tema degli esploratori: fiducia digitale e crittografia (elenco arXiv cs.CR, nuovi lavori del giorno).
Categorie arXiv: Cryptography and Security (cs.CR) ; Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.05453 — 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 2026-10-06T05:26:35.357Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).