aiqbench analysis
This survey provides a critical examination of security vulnerabilities within Retrieval-Augmented Generation (RAG) systems, specifically addressing threats like corpus poisoning and backdoor attacks that compromise the reliability of AI-generated content. By formalizing threat models across the full pipeline—from retrieval to generation—it offers essential insights for researchers aiming to harden LLM architectures against adversarial manipulation while maintaining factual grounding.
Retrieved But Not Reliable: A Survey on Attacks, and Defenses in Retrieval-Augmented Generation
arXiv:2608.24977v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) enhances large language models by grounding outputs in external knowledge, improving factuality and reducing hallucinations. At the same time, the retrieval-augmented pipeline introduces new robustness and security risks, including corpus poisoning, backdoor attacks, privacy leakage, and fairness violations. Despite rapid progress in this area, existing surveys remain limited in their treatment of attacker objectives, threat models, and stage-specific defenses across the full RAG pipeline. This survey presents a unified and pipeline-aware overview of RAG robustness. We formalize threat models over the corpus, retriever, and generator, and organize attacks into three main objectives: accuracy, privacy, and fairness. We further review defenses from a pipeline-aware perspective, covering the retrieval, rerank, generation, and traceback stages. In addition, we summarize robustness benchmarks and explainability methods for more deeply evaluating and explaining RAG robustness.
Authors:Minh Tran, Cuong Dang, Tuc Nguyen, Khanh-Tung Tran, Minh Huynh Nguyen, Trinh Chau, Kien Le, Do Xuan Long, Jiahao Zhang, Hoang D. Nguyen, Thanh Le, Suhang Wang
Abstract:Retrieval-Augmented Generation (RAG) enhances large language models by grounding outputs in external knowledge, improving factuality and reducing hallucinations. At the same time, the retrieval-augmented pipeline introduces new robustness and security risks, including corpus poisoning, backdoor attacks, privacy leakage, and fairness violations. Despite rapid progress in this area, existing surveys remain limited in their treatment of attacker objectives, threat models, and stage-specific defenses across the full RAG pipeline. This survey presents a unified and pipeline-aware overview of RAG robustness. We formalize threat models over the corpus, retriever, and generator, and organize attacks into three main objectives: accuracy, privacy, and fairness. We further review defenses from a pipeline-aware perspective, covering the retrieval, rerank, generation, and traceback stages. In addition, we summarize robustness benchmarks and explainability methods for more deeply evaluating and explaining RAG robustness.
24 pages, 6 figures. Accepted to Findings of the Association for Computational Linguistics: EMNLP 2026. Peer-reviewed through ACL Rolling Review (ARR)
Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
arXiv:2608.24977 [cs.CR]
(or arXiv:2608.24977v1 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2608.24977 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 206 words
Related coverage
Data Security in Large Language Models: Risks, Defense, and Directions
arXiv · Sep 16, 2026
DisarmRAG: Stealthy Retriever-Centric Poisoning to Disable Self-Correction in Retrieval-Augmented Generation (Extended Version)
arXiv · Jul 21, 2026
Digital States 2026: AI Moves From Experiment to Infrastructure
govtech.com · Sep 24, 2026
A Survey of Secure Retrieval-Augmented Generation
arXiv · Oct 8, 2026