TrustRAG: Enhancing Robustness and Trustworthiness in Retrieval-Augmented Generation

TL;DR

TrustRAG introduces a robust, training-free framework that enhances retrieval-augmented generation by filtering malicious content, improving accuracy, efficiency, and resistance to corpus poisoning attacks.

Contribution

It presents a novel two-stage filtering approach that systematically detects and removes malicious content before retrieval, improving the robustness of LLMs in RAG systems.

Findings

Significant improvement in retrieval accuracy.

Enhanced resistance to corpus poisoning attacks.

Seamless integration with existing LLMs.

Abstract

Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by integrating external knowledge sources, enabling more accurate and contextually relevant responses tailored to user queries. These systems, however, remain susceptible to corpus poisoning attacks, which can severely impair the performance of LLMs. To address this challenge, we propose TrustRAG, a robust framework that systematically filters malicious and irrelevant content before it is retrieved for generation. Our approach employs a two-stage defense mechanism. The first stage implements a cluster filtering strategy to detect potential attack patterns. The second stage employs a self-assessment process that harnesses the internal capabilities of LLMs to detect malicious documents and resolve inconsistencies. TrustRAG provides a plug-and-play, training-free module that integrates seamlessly with any open- or closed-source language model. Extensive experiments demonstrate that TrustRAG delivers substantial improvements in retrieval accuracy, efficiency, and attack resistance.