HyeonjeongHa/MM-PoisonRAG
Official PyTorch implementation of "MM-PoisonRAG: Disrupting Multimodal RAG with Local and Global Poisoning Attacks"
This project helps evaluate the security vulnerabilities of multimodal AI systems that answer questions using external knowledge. It takes an existing multimodal knowledge base (text and images) and a question-answering AI model. It then generates manipulated knowledge (text and images) designed to make the AI produce incorrect or nonsensical answers, allowing you to assess how easily an AI system can be misled. This is useful for AI security researchers, red teamers, or developers building multimodal RAG systems.
Use this if you need to understand and test how vulnerable your multimodal AI models and their knowledge bases are to targeted misinformation or broad disruption.
Not ideal if you are looking for a tool to defend against these attacks, as this project focuses solely on generating and evaluating the attacks themselves.
Stars
12
Forks
1
Language
Python
License
—
Category
Last pushed
Dec 04, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/rag/HyeonjeongHa/MM-PoisonRAG"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
LLAMATOR-Core/llamator
Red Teaming python-framework for testing chatbots and GenAI systems.
sleeepeer/PoisonedRAG
[USENIX Security 2025] PoisonedRAG: Knowledge Corruption Attacks to Retrieval-Augmented...
kelkalot/simpleaudit
Allows to red-team your AI systems through adversarial probing. It is simple, effective, and...
JuliusHenke/autopentest
CLI enabling more autonomous black-box penetration tests using Large Language Models (LLMs)
SecurityClaw/SecurityClaw
A modular, skill-based autonomous Security Operations Center (SOC) agent that monitors...