Learn › Exchange Online › 3 · Email authentication & deliverability

Outbound reputation and the high-risk pool

Your tenant's outbound mail shares infrastructure with the world's — until your behaviour gets it segregated. How Microsoft routes suspect mail, what one compromised mailbox costs, and the recovery sequence.

The pool model

Normal outbound leaves via shared healthy IP pools. Mail that LOOKS risky — outbound spam verdicts, anomalous volume — gets routed via the high-risk delivery pool: separate IPs whose reputation is, by design, sacrificial. Receivers treat high-risk-pool IPs accordingly. The system protects the healthy pools (and everyone else's delivery) from YOUR bad day; the cost is that your bad day compounds — spam flagged → risk pool → more rejections (the 5.7.708 family in the NDR concept when Microsoft itself starts restricting the sender/tenant).

The usual trigger is one compromised mailbox running the BEC starter kit: burst sending, then your legitimate mail starts bouncing from partners — the incident announces itself via deliverability before security tooling speaks.

The escalation ladder Microsoft runs

Individual sender: outbound-spam policy thresholds hit → the user is restricted from sending (blocked-senders list in Defender, alert raised — the per-user circuit breaker). Tenant-level: enough badness → tenant sending restrictions (the 5.7.7xx responses). Your controls on the same ladder: outbound spam policy limits (hourly recipient caps, forwarding controls — in the nightly scan), which you should TIGHTEN from defaults so the circuit breaker fires before reputation damage, not after.

The recovery sequence (order matters)

  1. Contain the account (revoke sessions, reset, kill the inbox rules and OAuth grants — BEC concept).
  2. Clear the restriction (unblock the sender in Defender once clean).
  3. THEN wait out pool recovery — reputation heals on Microsoft's timeline, not your escalation's; partner-side blocks may need their delisting flows.
  4. Post-incident: findings into outbound-policy thresholds and the auth-module records (SPF/DKIM intact throughout? attackers love sending as you WITH your own authentication).

What to watch (proofs)

  • Restricted entities: the Defender restricted-senders view + its alert — the circuit breaker's state; wire the alert to your channel (serv365 delivery pattern).
  • Outbound spam policy: Get-HostedOutboundSpamFilterPolicy (nightly scan) — thresholds and forwarding stance; drift here is security-posture drift.
  • The early smoke: outbound volume per sender trends (mail-flow reports) — the burst is visible BEFORE the bounces if anyone looks; alert on deviation.
  • Far-side evidence: 5.7.708-family and partner-bounce text trends (NDR ledger) — the reputation state as receivers see it.

PowerShell for this concept

Discussion

No messages yet — start the thread.

Sign in with your email to join the discussion — we send a one-time link, no password.