Enhanced Interactive Generative Model Agent for CTF Challenges

Talor Abramovich1, Meet Udeshi2, Minghao Shao2, Kilian Lieret3, Haoran Xi2, Kimberly Milner2, Sofija Jancheska2, John Yang4, Carlos E. Jimenez3, Farshad Khorrami2, Prashanth Krishnamurthy2, Brendan Dolan-Gavitt2, Muhammad Shafique5, Karthik Narasimhan3, Ramesh Karri2, and Ofir Press3

Although language model (LM) agents are demonstrating growing potential in many domains, their success in cybersecurity has been limited due to simplistic design and the lack of fundamental features for this domain. We present EnIGMA, an LM agent for autonomously solving Capture The Flag (CTF) challenges. EnIGMA introduces new Agent-Computer Interfaces (ACIs) to improve the success rate on CTF challenges. We establish the novel Interactive Agent Tools concept, which enables LM agents to run interactive command-line utilities essential for these challenges. Empirical analysis of EnIGMA on over 350 CTF challenges from three different benchmarks indicates that providing a robust set of new tools with demonstration of their usage helps the LM solve complex problems and achieves state-of-the-art results on the NYU CTF and Intercode-CTF benchmarks, managing to solve more than three times more challenges of NYU CTF benchmark compared to previous best agent (the NYU CTF agent).

Want to try it yourself and explore our new agent? We are completely open-source! You can try it out in the SWE-agent repository GitHub Repo stars, read our documentation and explore more about the research work in our paper.

Results

BenchmarkModel% Solved
NYU CTFEnIGMA w/ Claude 3.5 Sonnet13.5
EnIGMA w/ GPT-4 Turbo (1106)7.0
EnIGMA w/ GPT-4o9.0
NYU CTF agent w/ GPT-4 Turbo4.0
InterCode-CTFEnIGMA w/ Claude 3.5 Sonnet67.0
EnIGMA w/ GPT-4 Turbo (1106)72.0
EnIGMA w/ GPT-4o69.0
InterCode-CTF Agent40.0
Google DeepMind Agent w/ Gemini Ultra24.0
HackTheBoxEnIGMA w/ Claude 3.5 Sonnet26.0
EnIGMA w/ GPT-4 Turbo (1106)18.0
EnIGMA w/ GPT-4o16.0
NYU CTF agent w/ GPT-4 Turbo20.0

How it Works

Interactive Agent Tools In Action

BibTeX

If you found this work helpful, please consider using the following citation:

@misc{abramovich2024enigmaenhancedinteractivegenerative,
      title={EnIGMA: Enhanced Interactive Generative Model Agent for CTF Challenges},
      author={Talor Abramovich and Meet Udeshi and Minghao Shao and Kilian Lieret and Haoran Xi and Kimberly Milner and Sofija Jancheska and John Yang and Carlos E. Jimenez and Farshad Khorrami and Prashanth Krishnamurthy and Brendan Dolan-Gavitt and Muhammad Shafique and Karthik Narasimhan and Ramesh Karri and Ofir Press},
      year={2024},
      eprint={2409.16165},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2409.16165},
}
  1. Tel-Aviv University ↩︎

  2. New York University ↩︎ ↩︎2 ↩︎3 ↩︎4 ↩︎5 ↩︎6 ↩︎7 ↩︎8 ↩︎9

  3. Princeton Language and Intelligence, Princeton University ↩︎ ↩︎2 ↩︎3 ↩︎4

  4. Stanford University ↩︎

  5. New York University Abu Dhabi ↩︎


© 2024. All rights reserved.

Powered by Hydejack v9.2.1