paper-with-me

Papers

The Hidden Threat in Plain Text: Attacking RAG Data Loaders

2025-07-07 · Alberto Castagnaro, Umberto Salviati, Mauro Conti, Luca Pajola, Simeone Pizzi arxiv

Large Language Models (LLMs) have transformed human-machine interaction since ChatGPT's 2022 debut, with Retrieval-Augmented Generation (RAG) emerging as a key framework that enhances LLM outputs by integrating external knowledge. However, RAG's reliance on ingesting external documents introduces new vulnerabilities. This paper exposes a critical security gap at the data loading stage, where malicious actors can stealthily corrupt RAG pipelines by exploiting document ingestion. We propose a taxonomy of 9 knowledge-based poisoning attacks and introduce two novel threat vectors -- Content Obfuscation and Content Injection -- targeting common formats (DOCX, HTML, PDF). Using an automated toolkit implementing 19 stealthy injection techniques, we test five popular data loaders, finding a 74.4% attack success rate across 357 scenarios. We further validate these threats on six end-to-end RAG systems -- including white-box pipelines and black-box services like NotebookLM and OpenAI Assistants -- demonstrating high success rates and critical vulnerabilities that bypass filters and silently compromise output integrity. Our results emphasize the urgent need to secure the document ingestion process in RAG systems against covert content manipulations.

📄 PDF Abstract BibTeX arXiv:2507.05093

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Unveiling Hidden Pivotal Players with GoalNet: A GNN-Based Soccer Player Evaluation System

2025-03-12 · Jacky Hao Jiang, Jerry Cai, Anastasios Kyrillidis

Soccer analysis tools emphasize metrics such as expected goals, leading to an overrepresentation of attacking players' contributions and overlooking players who facilitate ball control and link attacks. Examples include …

Graph Attention

Rethinking Stealthiness of Backdoor Attack against NLP Models

2021-08-01 · ACL 2021 5 · Wenkai Yang, Yankai Lin, Peng Li, Jie zhou 외

Recent researches have shown that large natural language processing (NLP) models are vulnerable to a kind of security threat called the Backdoor Attack. Backdoor attacked models can achieve good performance on clean test…

Backdoor AttackData AugmentationSentiment AnalysisWord Embeddings

Adaptive Feature Alignment for Adversarial Training

2021-05-31 · Tao Wang, Ruixin Zhang, Xingyu Chen, Kai Zhao 외

Recent studies reveal that Convolutional Neural Networks (CNNs) are typically vulnerable to adversarial attacks, which pose a threat to security-sensitive applications. Many adversarial defense methods improve robustness…

Adversarial Defense

Invisible Safety Threat: Malicious Finetuning for LLM via Steganography

2026-03-09 · Guangnian Wan, Xinyin Ma, Gongfan Fang, Xinchao Wang arxiv

Understanding and addressing potential safety alignment risks in large language models (LLMs) is critical for ensuring their safe and trustworthy deployment. In this paper, we highlight an insidious safety threat: a comp…

LimeAttack: Local Explainable Method for Textual Hard-Label Adversarial Attack

2023-08-01 · Hai Zhu, Zhaoqing Yang, Weiwei Shang, Yuren Wu

Natural language processing models are vulnerable to adversarial examples. Previous textual adversarial attacks adopt gradients or confidence scores to calculate word importance ranking and generate adversarial examples.…

Adversarial AttackHard-label Attack