MyLLMAuto: A LLM CTF Challenge
MyLLMAuto is a Capture The Flag (CTF) application aimed at educating users about prompt injection vulnerabilities in multi-chain LLM applications. The application simulates an automotive parts lookup system with intentional security flaws, allowing users to explore and learn about security risks in LLM applications.
Key Features:
- Multi-Chain Architecture: Demonstrates the potential for prompt injection attacks across different LLM chains.
- Flag Submission System: Includes a UI for tracking and validating captured flags.
- Educational Focus: Designed for learning about security vulnerabilities, with intentional flaws included for demonstration purposes.
Benefits:
- Hands-On Learning: Users can engage with real-world security challenges in a controlled environment.
- Open Source: Contribute to the project and learn from the community.
- Docker Support: Easy setup with Docker and Docker Compose for local or containerized environments.
Highlights:
- Built using Python and FastAPI, leveraging OpenAI's GPT-3.5-turbo model.
- Includes multiple flags to discover through various techniques, enhancing the learning experience.