Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note2 min read

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

Source

arXiv

arxiv.org/abs/2606.24623v1

Read the source

Source

Plain-Language Brief

The authors propose a multi-agent framework that uses semantic rewriting to enhance large language models in sensitive scenarios while preserving privacy and contextual fidelity. The framework was tested on two datasets and six large language models, showing significant reduction in privacy leakage under targeted attacks compared to the existing SAGE method.

What the Document Says

  • The work introduces a multi-agent framework for Privacy-Preserving RAG (Retrieval-Augmented Generation).
  • The goal is to sanitize retrieved content through semantic rewriting in sensitive scenarios without compromising contextual fidelity.
  • Three specialized agents are employed: privacy extraction, semantic analysis, and reconstruction.
  • Experiments were conducted on ChatDoctor and Wiki-PII datasets across six large language models.
  • Results show a significant reduction in privacy leakage under targeted attacks.
  • For instance, LLaMA-3-8B's targeted information exposure reduced from 144 instances in the baseline to 1.
  • The framework achieves a BLEU-1 score of 0.122, outperforming SAGE method's 0.117.
  • The framework operates as an asynchronous preprocessing module with no additional latency during online inference.
  • The source code is publicly available at https://github.com/foursoils/Privacy-Preserving-RAG.

Key Concepts Mentioned

  • Retrieval-Augmented Generation
  • privacy preservation
  • multi-agent framework
  • semantic rewriting
  • ChatDoctor
  • Wiki-PII datasets
  • large language models
  • targeted attacks
  • BLEU score

Claims Made by the Document

  • The proposed multi-agent framework reduces privacy leakage under targeted attacks.
  • The framework outperforms the existing SAGE method in terms of BLEU-1 score.

Evidence Used

  • Results from experiments conducted on ChatDoctor and Wiki-PII datasets across six large language models.
  • Comparison between LLaMA-3-8B's targeted information exposure before and after applying the framework.

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.