model-extraction

name	model-extraction
version	2.0.0
description	Techniques to extract model weights, architecture, and training data through API queries
sasmp_version	1.3.0
bonded_agent	04-llm-vulnerability-analyst
bond_type	PRIMARY_BOND
input_schema	{"type":"object","required":["target_api"],"properties":{"target_api":{"type":"string"},"extraction_type":{"type":"string","enum":["query_based","distillation","embedding","architecture","all"]},"query_budget":{"type":"integer","default":10000}}}
output_schema	{"type":"object","properties":{"queries_used":{"type":"integer"},"fidelity_score":{"type":"number"},"extraction_success":{"type":"boolean"}}}
owasp_llm_2025	["LLM03","LLM02"]
mitre_atlas	["AML.T0024","AML.T0044"]

Model Extraction Attacks

Test AI systems for model theft vulnerabilities where attackers can reconstruct models through queries.

Quick Reference

Skill:       model-extraction
Agent:       04-llm-vulnerability-analyst
OWASP:       LLM03 (Supply Chain), LLM02 (Sensitive Info Disclosure)
MITRE:       AML.T0024 (Model Stealing)
Risk Level:  HIGH

Extraction Techniques

1. Query-Based Extraction

Technique: query_based
Queries Required: 10,000-100,000
Fidelity: 70-90%
Detection: Medium

Protocol:
  1. Generate diverse query set
  2. Collect model responses
  3. Train surrogate model
  4. Validate fidelity