model-extraction

What problem does it solve?

This Skill helps security teams evaluate how resilient AI models are to extraction attempts by simulating controlled probing and logging defenses.

Core Features & Use Cases

• Structured attack techniques: Provides query-based extraction, distillation, embedding theft, and architecture probing scenarios for risk assessment.
• Detection-focused evaluation: Includes indicators and metrics to measure fidelity, surrogate behavior, and defense effectiveness.
• Real-world use case: Run authorized security tests against deployed models to identify vulnerability patterns and strengthen safeguards.

Quick Start

Run the extraction test script to simulate attacks and generate a report.