Privacy-Preserving RAG via Multi-Agent Semantic Rewriting: Achieving Confidentiality Without Compromising Contextual Fidelity

Source

Source name: arXiv
Source kind: paper
Authors / organization: Yuanhe Zhao, Tianyu Zhang, Huafei Xing, Derek F. Wong, Jianbin Li, Tao Fang
Venue: arXiv: cs.CL
Published: 2026-06-23
Collected: 2026-06-24
URL: http://arxiv.org/abs/2606.24623v1
PDF: https://arxiv.org/pdf/2606.24623v1
DOI: UNKNOWN

Plain-Language Brief

Researchers propose a new method to enhance large language models while maintaining privacy in sensitive scenarios. The method uses a multi-agent framework that sanitizes retrieved content through semantic rewriting, reducing privacy leakage and maintaining contextual fidelity. The results of the experiment show a significant reduction in privacy leakage under targeted attacks compared to the existing SAGE method, without introducing additional latency.

What the Document Says

Retrieval-Augmented Generation (RAG) enhances large language models by incorporating external knowledge. However, deploying it in sensitive scenarios risks privacy leakage via malicious prompts.
The researchers propose a multi-agent framework that sanitizes retrieved content through semantic rewriting.
The framework consists of three specialized agents: for privacy extraction, semantic analysis, and reconstruction.
The approach collaboratively removes sensitive identifiers while preserving the semantic core.
The researchers evaluate the framework on ChatDoctor and Wiki-PII datasets across six large language models.
Experimental results demonstrate a significant reduction in privacy leakage under targeted attacks.
For example, they reduced targeted information exposure in LLaMA-3-8B from 144 instances in the baseline to just 1.
The framework maintains strong contextual fidelity with a BLEU-1 score of 0.122, outperforming the existing SAGE method's 0.117.
The framework operates as an asynchronous preprocessing module, introducing no additional latency to online inference.
All rewriting is executed as a one-time offline preprocessing step.
The source code of this work is publicly available at https://github.com/foursoils/Privacy-Preserving-RAG.

Key Concepts Mentioned

Retrieval-Augmented Generation
privacy leakage
multi-agent framework
semantic rewriting
sensitive scenarios
ChatDoctor
Wiki-PII datasets
large language models
targeted attacks
contextual fidelity
BLEU-1 score
SAGE method
asynchronous preprocessing module
offline inference

Claims Made by the Document

The proposed method significantly reduces privacy leakage under targeted attacks.
The framework maintains strong contextual fidelity.

Evidence Used

Experimental results demonstrate a significant reduction in privacy leakage...
The framework maintains strong contextual fidelity with a BLEU-1 score of 0.122...

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Retrieval-Augmented Generation enhances large language models by incorporating external knowledge, but deploying it in sensitive scenarios risks privacy leakage via malicious prompts. To address this, we propose a multi-agent framework that sanitizes retrieved content through semantic rewriting. By employing three specialized agents for privacy extraction, semantic analysis, and reconstruction, our approach collaboratively removes sensitive identifiers while preserving the semantic core. We evaluate the framework on the ChatDoctor and Wiki-PII datasets across six large language models. Experimental results demonstrate a significant reduction in privacy leakage under targeted attacks. For instance, we reduced targeted information exposure in LLaMA-3-8B from 144 instances in the baseline to just 1. Furthermore, we maintain strong contextual fidelity with a BLEU-1 score of 0.122, outperforming the existing SAGE method's 0.117. Finally, the framework operates as an asynchronous preprocessing module, introducing no additional latency to online inference, as all rewriting is executed as a one-time offline preprocessing step. To promote reproducibility, the source code of this work is publicly available at https://github.com/foursoils/Privacy-Preserving-RAG.