跳到正文
HuggingFace Daily Papers(社區熱門論文)·· 18 天前AI 評分52

簽署後仍可操控:衡量並縮小離線 RAG 的文字與向量完整性差距

Signed but Steerable: Measuring and Closing the Text-Vector Integrity Gap in Offline Retrieval-Augmented Generation

AI 導讀

該研究衡量離線 RAG 簽署知識包的文字與向量完整性差距,並比較向量重算、壓縮索引審核及量化碼驗證方法。理想化浮點噪聲級容差仍令最多 5.5% 查詢暴露;在包含 57,638 份文件的 FAISS 乘積量化索引中,通過審核的篡改向量可令攻擊者文件在 25% 目標中排首位。改為驗證量化碼並加入校準保護區及規範化後,三個平台的樣本中所有誠實向量均獲接受。

正文

Published on Nov 4

Authors:

Abstract

Signed retrieval packs can contain embedding vectors that do not match their text, and practical vector verification and compressed-index audits leave many queries vulnerable, while code-level verification with calibrated ambiguity handling accepts honest vectors.

Offline retrieval-augmented generation (RAG) on edge devices is updated through signed knowledge packs that carry text and precomputed embedding vectors. A signature proves who published a pack, not that its vectors encode its text. We study this text-vector integrity gap. A verifier that recomputes vectors and accepts them within a cosine tolerance concedes a predictable share of queries to an attacker who controls the vectors: the hijack rate equals the distribution of angular retrieval margins at the tolerance (the margin law). Margins are small in five corpora and two embedding models, so an idealised tolerance at the level of floating-point noise leaves up to 5.5% of queries exposed (at most 1.2% in any corpus once exact duplicate documents are excluded), and a threshold calibrated from our CPU INT8-FP32 comparison of a small embedding model (at the empirical 0.1% quantile of honest cosines) leaves 98-100% exposed; on a Raspberry Pi 5 running a local language model, such an attack lowers correct answers from 0.75 to 0.39. Integrity audits of compressed vector stores fare no better: on a FAISS product-quantisation index over 57,638 documents, tampered vectors that all pass the audit put the attacker's document first for 25% of targets. Verifying quantisation codes instead, with ambiguity decided by the verifier, a calibrated guard band and canonicalisation, accepted every honest vector in our samples on three platforms and left the publisher no choice. We also show that common floating-point and dynamic-INT8 inference pipelines are not bit-reproducible across batch sizes or architectures, and report pre-registered negative results on abstention, context pruning and semantic screening. We release code and data.

View arXiv page View PDF Project page GitHub Add to collection

Get this paper in your agent:

hf papers read 2611.06498

Don't have the latest CLI?

curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2611.06498 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2611.06498 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2611.06498 in a Space README.md to link it from this page.

Collections including this paper 1

來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co