Risk Detection Rules

The compliance engine uses behavioral heuristics + ML inference to detect high-risk activity without accessing identity data.

🛰️ Types of Signals Analyzed

Category	Examples	Privacy Exposure
Liquidity Behavior	Sudden collateral movement, cycling	No identity revealed
Mint / Burn Patterns	Rapid mint/redeem loops	No identity revealed
Transaction Graph Shape	Dense or circular flow networks	Uses hashed graph state
Cross-Boundary Patterns	Bridges to sanctioned zones	Region-only metadata, no address insight

🔥 Risk Scoring Model

Risk Score = Behavioral Outlier Index + Liquidity Volatility + Repeat Pattern Weight

Scores exceed 75 → flagged to DAO review queue Scores exceed 90 → automatic mint throttle

🧱 Hard Stops

The system cannot:

• Freeze wallets

• Seize assets

• Identify users

• Reverse transactions

It only adjusts protocol parameters and forwards risk alerts to the DAO.