Security.md

Security

Flashix security is built around layered controls, explicit approvals, and auditability.

Security Model

• Sealed compute prevents untrusted hosts from fabricating signals.
• Canonical signal encoding prevents verifier drift.
• Signature recovery on-chain rejects forged execution instructions.
• Flashloan repayment checks prevent partial settlement.
• Circuit breakers stop execution when market or operational conditions degrade.
• Human overrides provide a last-resort emergency control.

Key Management

• Never commit deploy keys or TEE secrets.
• Keep local .env files out of source control.
• Rotate API keys and signing keys when provisioning changes.
• Use the sealed keystore path for inference signing and verify the recovered address after load.

Execution Safety

• Simulation must happen before broadcast.
• Signals expire quickly and are rejected if stale.
• Replay protection blocks reused opportunity IDs.
• Router allowlists prevent arbitrary DEX execution.
• Profit floors and gas thresholds prevent low-value trades.

Operational Checks

• Run the pre-deployment checklist before contracts go live.
• Verify the backend /health endpoint before demoing.
• Confirm the frontend is connected to the intended backend.
• Ensure the persistence service is running before generating exports.

Evidence To Keep

• Deployment manifests.
• Explorer links.
• Decision logs.
• Ledger exports.
• Replay validation reports.