← Back to Database
Behavioural Model Fingerprinting
🟡 MEDIUM
model_extraction
proven
AVE-2025-0067
· aka: Model ID via Behaviour
Summary
Systematic probing of agent responses reveals the underlying model type, version, and configuration.
Blast Radius
Attacker can tailor exploits to the specific model.
Prerequisites
Access to agent API.
Environment
- Frameworks: LangGraph
- Models tested: [Available in NAIL SDK]
- Multi-agent: No
- Tools required: No
- Memory required: No