Dataset and Lessons Learned from the 2024 SaTML LLM Capture-the-Flag Competition

TL;DR

This paper presents a dataset and insights from a Capture-the-Flag competition on LLM security, revealing the challenges in defending against malicious prompts and the need for further research.

Contribution

It introduces a large dataset of attack chats, analyzes the competition results, and highlights the difficulty of creating foolproof defenses for LLMs.

Findings

All defenses were bypassed at least once

The competition revealed the complexity of securing LLMs

A dataset of 137k attack chats was released

Abstract

Large language model systems face important security risks from maliciously crafted messages that aim to overwrite the system's original instructions or leak private data. To study this problem, we organized a capture-the-flag competition at IEEE SaTML 2024, where the flag is a secret string in the LLM system prompt. The competition was organized in two phases. In the first phase, teams developed defenses to prevent the model from leaking the secret. During the second phase, teams were challenged to extract the secrets hidden for defenses proposed by the other teams. This report summarizes the main insights from the competition. Notably, we found that all defenses were bypassed at least once, highlighting the difficulty of designing a successful defense and the necessity for additional research to protect LLM systems. To foster future research in this direction, we compiled a dataset with over 137k multi-turn attack chats and open-sourced the platform.