What question did this study set out to answer?

The main aim is to present RAG Shield, a defense framework for protecting RAG pipelines from poisoning and adversarial attacks.

February 2, 2026Open Access

RAG Shield: A Multi-Layer Defense System Against Poisoning Attacks in Retrieval-Augmented Generation

Key Points

The main aim is to present RAG Shield, a defense framework for protecting RAG pipelines from poisoning and adversarial attacks.
Introduces a multi-layer defense architecture
Combines document provenance validation and semantic anomaly detection
Evaluates the system against various attack scenarios in controlled settings
Establishes a realistic threat model
Demonstrates the effectiveness of the defense system against multiple poisoning attack scenarios
Highlights the role of document provenance and anomaly detection in enhancing security
Supports open dialogue on AI security and secure enterprise RAG deployment

Abstract

This whitepaper presents RAG Shield, a security-focused framework fordefending Retrieval-Augmented Generation (RAG) pipelines againstpoisoning and adversarial manipulation. The work introduces a multi-layer defense architecture combiningdocument provenance validation, semantic anomaly detection, andsecure retrieval mechanisms. A realistic threat model is defined,and the system is evaluated against multiple attack scenarios undercontrolled conditions. This document is released as a technical preprint to establishprior art and support open discussion in the areas of AI security,adversarial machine learning, and secure enterprise RAG deployment.

Read Full Paperexternally

Bookmark

View Full Paper

Cite This Study

Fabio Petti (Sun,) studied this question.

synapsesocial.com/papers/6980ffd6c1c9540dea812b31 https://doi.org/https://doi.org/10.5281/zenodo.18449664

Bookmark

View Full Paper