Characterising Indirect Prompt Injection in Retrieval-Augmented Generation Systems
My current paper on indirect prompt injection as a context-integrity problem, with a threat model, taxonomy, and evaluation framing for RAG systems.
Research
Writing and research around systems, security, and AI.
Same design language as the blog, but each entry now goes directly to the actual research paper.
LLM-Firewall: RAG Detection Middleware Architecture
Personal research on the RAG detection middleware architecture I designed to detect and sanitize IPI attacks in real-time.
LLM-Firewall: A Novel Taxonomy of IPI Attacks
A Novel Taxonomy of Indirect Prompt Injection Attacks in Enterprise RAG Systems.