This research introduces a RAG-based LLM architecture enhanced with chain-of-thought reasoning to generate higher-quality network protocol test seeds for security fuzzing.

• Combines retrieval-augmented generation with text embeddings in a two-stage approach
• Leverages chain-of-thought prompting to improve structural quality of generated test packets
• Enables more comprehensive exploration of protocol state spaces for vulnerability discovery
• Demonstrates how AI agents can improve cybersecurity testing methodologies

Significant for security teams as it presents a novel approach to automate and enhance the quality of network protocol testing, potentially uncovering vulnerabilities that traditional methods might miss.

Retrieval Augmented Generation Based LLM Evaluation For Protocol State Machine Inference With Chain-of-Thought Reasoning