Updated on
December 19, 2024
Model Leeching: An Extraction Attack Targeting LLMs
Model Leeching: An Extraction Attack Targeting LLMs
Dr. Peter Garraghan
TABLE OF CONTENTS
Key Takeaways
Key Takeaways

Model Leeching is a novel extraction attack targeting Large Language Models (LLMs), capable of distilling task-specific knowledge from a target LLM into a reduced parameter model.

     

DOWNLOAD RESEARCH

 

We demonstrate the effectiveness of our attack by extracting task capability from ChatGPT-3.5-Turbo, achieving 73% Exact Match (EM) similarity, and SQuAD EM and F1 accuracy scores of 75% and 87%, respectively for only $50 in API cost.

We further demonstrate the feasibility of adversarial attack transferability from an extracted model extracted via Model Leeching to perform ML attack staging against a target LLM, resulting in an 11% increase to attack success rate when applied to ChatGPT-3.5-Turbo.

Access the complete insights into Model Leeching.

Next Steps

Thank you for reading our research about Model Leeching!

  1. Test Our Free Platform: Experience how our Automated Red Teaming platform swiftly identifies and remediates AI security vulnerabilities. Start for free today!
  2. Follow Mindgard: Stay updated by following us on LinkedIn and X, or join our AI Security community on Discord.
  3. Get in Touch: Have questions or want to explore collaboration opportunities? Reach out to us, and let's secure your AI together.
  4. Please, feel free to request a demo to learn about the full benefits of Mindgard Enterprise.
blog icon
What happens when you turn an AI model’s safety controls off?

Explore how model abliteration removes AI safety controls, the legitimate security applications it enables and the risks unrestricted models create for cyberattacks, evaluation and regulation.

blog icon
Research: Shadow AI is a Blind Spot in Enterprise Security, Including Among Security Teams

At RSA Conference 2025 and InfoSecurity Europe 2025 we surveyed over 500 cybersecurity professionals to assess emerging threats in enterprise environments. The findings reveal a growing and often overlooked risk: security professionals using generative AI tools without approval, a trend known as Shadow AI.

blog icon
Who Guards the Guardrails? Five Questions AI Security Buyers Should Ask

Understand AI guardrail and gateway weaknesses, how to scrutinize and interpret reported accuracy claims, as well as provide actionable guidance on how vendors can more independently assess and ensure their guardrail's effectiveness.

Start Securing Your AI Systems

See how Mindgard exposes and fixes exploitable AI risk across your AI agents and systems.

Book a Demo

Mindgard helps organizations discover, assess and defend their AI systems. Spun out of more than a decade of AI security research at Lancaster University in the UK and headquartered in Boston and London, Mindgard operationalizes the expertise of AI researchers and offensive security practitioners through a Security Platform that performs Shadow AI discovery, AI red teaming, and run-time AI protection to assess and mitigate risk across models, agents, and applications.

AICPA SOC SOC 2 Type 2 Compliant