[ Transport security ]

MTA-STS testing and enforce modes: choose a safe transition

Move from testing to enforce only after identifying transport failures and proving that all intended receiving routes remain usable.

By the EmailAuth research team

[ Key points ]

  • Testing can permit delivery despite a policy failure.
  • Check backup routes before promoting to enforce.
  • New policy announcements do not instantly replace all cached versions.
On this page

Testing and enforce use the same MTA-STS policy checks but attach different delivery consequences. Testing is an observation stage. Enforce tells supporting senders not to deliver to MX hosts that fail the policy's host, STARTTLS or certificate checks. Neither mode replaces SPF, DKIM or DMARC.

Establish what testing can reveal

References: RFC 8461 §5, RFC 8460 §3.

Publish a valid testing policy that lists the intended receiving MX hosts, and enable TLS-RPT with an owned reporting destination. Senders implementing both mechanisms can report policy failures while still delivering as though no MTA-STS validation failure occurred. Successful inbox delivery in testing therefore does not prove that the same route will work in enforce.

Record the current MX set, certificate names, expiration dates, supported TLS configuration and fallback routes. Inspect each candidate host, including higher-preference-number backups that may receive little normal traffic. Reports describe participating senders' observations, not every message and not an exhaustive internet-wide health scan.

Resolve failures by their actual cause

References: RFC 8461 §3.3, RFC 8461 §4.1, RFC 8461 §4.2.

An MX-pattern mismatch calls for checking the intended routing inventory and policy. A certificate-name failure calls for checking the exact MX hostname against its presented certificate. A policy-fetch error belongs to the HTTPS host. Do not respond to all three by broadening the wildcard or turning off authentication elsewhere.

For each issue, preserve reporter, period, affected host and evidence before making the smallest approved change. Confirm the repair with an explicit probe and later reporting where available. Lack of new reports alone is not proof of repair, particularly for low-volume or intermittently used backup routes.

Work through an enforce rehearsal

References: RFC 8461 §5, RFC 8461 §5.1.

Use a synthetic domain whose policy permits mx1.example.net and mx2.example.net. Give mx1 valid STARTTLS and a matching trusted certificate; configure the lab's mx2 to present an unrelated certificate. In testing, the second route represents a reportable policy failure but MTA-STS does not require withholding delivery. In enforce, a supporting sender must not deliver to that failing candidate and may try another candidate.

Repeat with both candidates invalid and inspect the sender queue and retry logs. RFC 8461 recommends treating these failures as transient and requires checking for an updated policy before permanently failing delivery. This is a reproducible lab procedure; no live MTA execution or queue observation was performed for this article.

Define a promotion and reversal decision

References: RFC 8461 §8.1, RFC 8461 §8.3.

Promote only with owned evidence for all legitimate routes, a functioning reporting path, a documented support escalation and a tested policy-edit workflow. Publish the HTTPS body before changing the DNS id. Account for both DNS caching and the previously published policy max_age. A new testing policy does not instantly replace every cached enforce policy, so restoring compliant service may be the fastest recovery action.

Retained historical evidence and remaining scope

The following is the retained 7 November 2026 lab record, inspected but not rerun as a network experiment for this review. New offline fixture execution is recorded separately.

The executed loopback trace queried _mta-sts.example.test TXT over UDP and received v=STSv1; id=A1 with TTL 300. A separate MX query returned mx.old.example.test. HTTPS then returned a 200 policy body in testing mode, and an SMTP EHLO advertised STARTTLS. The first handshake succeeded with TLS 1.3 and a certificate trusted through the disposable lab CA. Replacing the SMTP certificate with one for wrong.example.test produced a hostname-verification failure. The restricted mode-decision fixture allowed the testing case and withheld the enforce candidate. No email was transmitted, and no actual MTA queue was exercised. This proves the controlled DNS, policy-fetch and TLS boundary, while production route coverage and sender implementation testing remain pending.

[ Takeaway ]

New policy announcements do not instantly replace all cached versions.

[ By role ]

What this means for you

Choose your role for practical next steps.

IT / Mail Admin

Observe the quiet backup routes

Normal traffic can miss a broken fallback MX.

Next steps

  1. Test backup MX hosts.
  2. Correlate reports with probes.
  3. Rehearse queued delivery failures.

CISO / SecOps

Bound the enforcement promise

Only participating senders apply this transport mechanism.

Next steps

  1. Distinguish transport from sender authentication.
  2. Review all policy failure categories.
  3. Preserve evidence before broadening policy.

Founder / Non-technical

Promote with route coverage

Enforcement changes the consequences of failures that testing may tolerate.

Next steps

  1. Approve a coverage-based promotion gate.
  2. Keep service owners available.
  3. Budget for delayed-mail recovery.

[ FAQs ]

Frequently asked questions

Does testing block noncompliant delivery?

Not because of the MTA-STS validation failure; other sender policies can still block it.

RFC 8461 §5

Does enforce guarantee every sender uses TLS?

No. The behavior applies to senders that implement and apply MTA-STS.

RFC 8461 §1

How many days should testing last?

Choose based on traffic and route coverage, not a universal fixed number.

Can reports replace direct probes?

No. Test low-volume and backup routes explicitly.

Must every MX be in the policy?

Every intended receiving candidate should match an allowed pattern.

RFC 8461 §4.1

Is a clean inbox test enough?

No. Testing may allow delivery despite validation failure.

RFC 8461 §5

What if every MX fails in enforce?

Supporting senders must not deliver through those failures and normally retry.

RFC 8461 §5.1

Can I reverse by deleting DNS?

That does not remove valid cached policies. Use a planned policy update.

RFC 8461 §8.3

[ Sources ]

Sources

  1. RFC 8460

    Sections §1 · §3

  2. RFC 8461

    Sections §1 · §3.3 · §4.1 · §4.2 · §5 · §5.1 · §8.1 · §8.3

[ Glossary ]

Glossary

Testing
Mode permitting delivery despite MTA-STS policy validation failures.

RFC 8461 §5

Enforce
Mode forbidding delivery to candidates that fail applicable checks.

RFC 8461 §5

TLS-RPT
Aggregate reporting of SMTP TLS policy outcomes.

RFC 8460 §1

Enterprise provenance

Why top financial institutions and governments retired their email security dashboards.

Traditional email authentication platforms sell you a dashboard and ask your engineers to do the labour. In high-stakes environments, manual oversight means delayed enforcement, configuration drift and unmonitored shadow domains.

EmailAuth is the agentic email provenance platform. Our agents enforce p=reject, keep third-party senders aligned and move your domains to full RFC 9989 compliance, the new DMARC standard,, with human-in-the-loop auditability.

Governments Large enterprises Tier-1 financial institutions

SaaS · Dedicated tenancy · Air-gapped on-prem · BYOK encryption

Knowledge Base