MIT01 — Agent identity authentication: every sender must present a verified credential before the channel is established or the message is accepted.

MIT03 — Encrypted channels (mTLS): messages are encrypted in transit, and the channel itself is mutually authenticated — combining confidentiality (stops Type 5) with identity proof (contributes to stopping Types 1, 3).

MIT06 — Least-privilege routing: infrastructure enforces that only authorized senders can reach authorized recipients, regardless of what credentials the sender presents.

MIT07 — Agent registry: the authoritative source of truth for which credentials are valid and which agents are authorized to communicate with which.

A message that doesn't pass Ring 1 never reaches the receiving agent at all.

MIT02 — Message signing: the receiving agent verifies the signature before processing — detecting any tampering between the authenticated sender and this point (stops Type 2).

MIT04 — Input validation: strict schema enforcement rejects malformed, out-of-scope, or attacker-extended messages before they reach the agent's reasoning.

MIT05 — Replay protection: nonce + timestamp validation rejects messages that were valid in the past but are not authorized for this moment (stops Type 4).

A message that passed Ring 1 but fails Ring 2 is rejected before the agent acts on it.

MIT08 — Audit logging: every message is logged with sender identity and content hash, providing post-incident reconstruction and enabling detection of credential misuse by legitimate agents.

MIT09 — Anomaly detection: behavioral baselines expose coordinated attacks and compromised agents through deviation alerting, even when individual messages pass all verification checks.

Minimum viable set: For most pipelines, MIT01 + MIT03 + MIT02 + MIT05 addresses the most critical risks. Add MIT06 for queue-based architectures and MIT07 as pipeline complexity grows.