StruQ

StruQ Overview

StruQ is an innovative solution designed to defend against prompt injection attacks targeting large language models (LLMs). It employs a structured query approach that separates trusted prompts from user-provided data, effectively mitigating manipulation risks.

Key Features

• Structured Queries: Enables a clear distinction between prompts and data, enhancing security.
• Fine-Tuning Strategy: Trains LLMs to ignore potentially harmful instructions embedded within user data.
• Enhanced Model Robustness: Significantly improves resistance to prompt injection attacks without sacrificing performance utility.

Benefits

• Security: Protects applications from the prevalent threat of prompt injection attacks.
• Usability: Retains high utility for applications leveraging LLM capabilities.
• Flexibility: Accommodates various implementations through its structured framework.

Highlights

• The system includes a secure front-end for formatting queries and a specially trained LLM for optimal output generation.
• Users can easily clone the repository and set up the environment for training and testing models.