analiza aiqbench
This survey systematically catalogs 135 studies on securing Retrieval-Augmented Generation (RAG) systems, addressing unique vulnerabilities introduced by external knowledge access paths. By introducing the SLOT taxonomy, it highlights critical gaps where attacks outpace defenses and proposes directions for standardized evaluation in multimodal and agentic contexts. The work provides essential metadata and labels for researchers aiming to harden LLM architectures against retrieval-specific threats.
A Survey of Secure Retrieval-Augmented Generation
arXiv:2604.08304v4 Announce Type: replace-cross Abstract: Retrieval-augmented generation (RAG) improves large language models (LLMs) with external knowledge, but this access path creates security risks distinct from inherent prompt-only or parametric-model flaws. We frame secure RAG as securing external knowledge access. We conducted a systematic search and curated 135 works on attacks, defenses, and security evaluation, and organized them with SLOT: a taxonomy along the attack Surface (S) and the corresponding defense Layer (L), with cross-cutting axes Objective (O) and attack-target scope (T). Mapping these studies onto an external knowledge-access pipeline, we expose three mismatches: target mismatch (T2 attacks outpace T2 defenses and evaluation), stage mismatch (S1 attacks outnumber L1 defenses), and signal mismatch (fluent, retrievable, corpus-fitting S1 attacks challenge anomaly-based L2/L3 defenses). Finally, we discuss directions for more realistic targets, surface-complete defense stacks, standardized evaluation, confidentiality, and multimodal and agentic systems, and release the screening process, selected-paper metadata, and machine-readable SLOT labels at https://github.com/TreeAI-Lab/Awesome-RAG-Security.
Abstract:Retrieval-augmented generation (RAG) improves large language models (LLMs) with external knowledge, but this access path creates security risks distinct from inherent prompt-only or parametric-model flaws. We frame secure RAG as securing external knowledge access. We conducted a systematic search and curated 135 works on attacks, defenses, and security evaluation, and organized them with SLOT: a taxonomy along the attack Surface (S) and the corresponding defense Layer (L), with cross-cutting axes Objective (O) and attack-target scope (T). Mapping these studies onto an external knowledge-access pipeline, we expose three mismatches: target mismatch (T2 attacks outpace T2 defenses and evaluation), stage mismatch (S1 attacks outnumber L1 defenses), and signal mismatch (fluent, retrievable, corpus-fitting S1 attacks challenge anomaly-based L2/L3 defenses). Finally, we discuss directions for more realistic targets, surface-complete defense stacks, standardized evaluation, confidentiality, and multimodal and agentic systems, and release the screening process, selected-paper metadata, and machine-readable SLOT labels at this https URL.
Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
arXiv:2604.08304 [cs.CR]
(or arXiv:2604.08304v4 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2604.08304 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 246 cuvinte
Subiecte conexe
Retrieved But Not Reliable: A Survey on Attacks, and Defenses in Retrieval-Augmented Generation
arXiv · 27 aug. 2026
Data Security in Large Language Models: Risks, Defense, and Directions
arXiv · 16 sept. 2026
Mapping the RAG Landscape: A Four Axis Taxonomy of Efficiency, Defense, Interactivity, and Reasoning
arXiv · 2 oct. 2026
A shared agentic platform for Wood Mackenzie, on Amazon Bedrock AgentCore
AWS · 17 sept. 2026