Software Development
Detecting Model Extraction Attacks
The skill monitors inference‑API usage to detect model extraction attacks, including stealing, inversion, and membership inference. By tracking per‑principal query volume, distribution, and rate, it applies rate limiting and output perturbation to mitigate risk. It is ideal for public or partner APIs and for pre‑deployment red‑team assessments of extraction vulnerability.
- Publisher / source
- Cybersecurity Skills (Community)
- Upstream record
- mukul975/Anthropic-Cybersecurity-Skills
- Version
- Version 3
- Published
- Publication review
- Approved for publication
- Content fingerprint
f721213e5733b63c2f6f3f26c134a3b8de96d75dc515600c0854fe97d223f4bd