Metraipolis › archive › dc_07442013c792

Transferable End-to-End Optimization for Indirect Long-Term Memory Poisoning in LLM Agents — arXiv:2609.00523

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.00523
License (as archived)
CC0-1.0
Archived
2026-10-17T06:44:00.296Z
SHA-256 of the archived text
f8c0f7c3007f2abe2fda05ead909ca659433a67226bd3aaecf949ae792cb4d8e
Stable URL of this page
https://metraipolis.com/documento/dc_07442013c792

How to cite

Transferable End-to-End Optimization for Indirect Long-Term Memory Poisoning in LLM Agents — arXiv:2609.00523 — arXiv (en), https://arxiv.org/abs/2609.00523 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_07442013c792 (Metraipolis archive dc_07442013c792, sha256 f8c0f7c3007f2abe2fda05ead909ca659433a67226bd3aaecf949ae792cb4d8e)
BibTeX
@misc{dc07442013c792,
  title = {Transferable End-to-End Optimization for Indirect Long-Term Memory Poisoning in LLM Agents — arXiv:2609.00523},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.00523},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_07442013c792; sha256 f8c0f7c3007f2abe2fda05ead909ca659433a67226bd3aaecf949ae792cb4d8e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.00523. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Long-term memory can turn untrusted external content into persistent influence over an LLM agent's future decisions, creating the threat of indirect memory poisoning. A successful attack must survive a multi-stage pipeline comprising memory writing, retrieval, and utilization. Existing attacks largely rely on intra-stage optimization, optimizing individual stages in isolation while overlooking inter-stage coupling. Specifically, these stages impose different requirements on the same poisoning content, and each stage operates on the transformed output of its predecessor. Consequently, optimizing one stage may undermine the effectiveness of other stages, while upstream transformations may erase improvements intended for downstream stages. Indirect memory poisoning should therefore be viewed as an end-to-end optimization problem. Based on this insight, we present \textsc{PipePoison}, which collects fine-grained stage feedback from local shadow systems, uses chain-structured losses to identify and optimize the stage bottlenecking end-to-end success, and applies stability-calibrated stage and configuration weights to improve transferability. Across three agent frameworks and four memory mechanisms, \textsc{PipePoison} improves attack utilization rate by 19.1 percentage points. Even on fully unseen victim configurations, it outperforms the strongest baseline by 16 percentage points and remains effective under eight representative defenses.

Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-02 in cs.CR).
Categorie arXiv: Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2609.00523 — 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 2026-10-17T06:38:14.356Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).