Metraipolis › archive › dc_64cafef8b3d9
Only Pay What You Must Spend: On-Demand Privacy Budget Payment for Differentially Private RAG — arXiv:2609.27406
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.27406
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T10:08:59.842Z
- SHA-256 of the archived text
cdb218a594d533066e209f26f74102c09a505aa3409bf9878d9429d0bb09959a- Stable URL of this page
https://metraipolis.com/documento/dc_64cafef8b3d9
How to cite
Only Pay What You Must Spend: On-Demand Privacy Budget Payment for Differentially Private RAG — arXiv:2609.27406 — arXiv (en), https://arxiv.org/abs/2609.27406 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_64cafef8b3d9 (Metraipolis archive dc_64cafef8b3d9, sha256 cdb218a594d533066e209f26f74102c09a505aa3409bf9878d9429d0bb09959a)
BibTeX
@misc{dc64cafef8b3d9,
title = {Only Pay What You Must Spend: On-Demand Privacy Budget Payment for Differentially Private RAG — arXiv:2609.27406},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.27406},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_64cafef8b3d9; sha256 cdb218a594d533066e209f26f74102c09a505aa3409bf9878d9429d0bb09959a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.27406. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Deploying large language models (LLMs) on sensitive data via Retrieval-Augmented Generation (RAG) introduces severe privacy risks. Recent studies apply Differential Privacy (DP) to LLMs with RAG for formal privacy guarantees. However, existing DP-RAG frameworks rapidly exhaust the privacy budget. Although recent efforts attempt to save the budget by narrowing the retrieval scope or sparsifying private generation, these methods themselves cumulatively consume the budget, whereas they could actually rely merely on public information or at a negligible one-time privacy cost. This mismatch fails to align budget expenditure with the model's actual reliance on private data, causing substantial waste on operations that require no private access. To address this, we propose SparsePay-RAG, adopting "only pay what you must spend" as its core principle. Using public information as a zero-privacy prior, it charges the privacy budget only for the private increment. Specifically, SparsePay-RAG narrows the retrieval scope via public topic-guided clustering, adaptively controls private access frequency without privacy cost through isotonic cross-layer trajectory fitting, and compresses per-access budget via DP contrastive decoding. Under strong privacy constraints, experiments show SparsePay-RAG achieves superior privacy-utility trade-offs over baselines. Tema degli esploratori: ricerca e scienza aperta (retro-catalogo arXiv: annunci del 2026-09-24 in cs.DL). Categorie arXiv: Cryptography and Security (cs.CR) ; Digital Libraries (cs.DL). Pagina del lavoro: https://arxiv.org/abs/2609.27406 — 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 2026-10-08T09:18:14.486Z dalla pagina https://arxiv.org/catchup/cs.DL/2026-09-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).