mailskop.

Guide ·

DMARC from p=none to p=reject, step by step

By: Kemal Özleyen, Cybersecurity Engineer · Check Point Workspace Security Expert · Kinetik Bilişim

p=none monitors but does not protect: forged mail is still delivered. RFC 9989 §5.1 describes confirming every legitimate sender with reports, then moving to quarantine and reject; you can test each step one level lower with pct=0; t=y.

Correction (): In an earlier version the test steps carried only t=y. Receivers working to RFC 7489 do not recognise t, so they could enforce those records in full; the steps now use pct=0; t=y.

Check your record now

Why p=none is not enough

RFC 9989 defines p=none as “Monitoring Mode”: it expresses no preference to the receiver and only asks for reports. Mail impersonating your domain is delivered even though it fails authentication.

In Mailskop’s September 2026 measurement, 24.1% of .tr domains were at p=none; 8.9% enforced reject.

The steps

  1. Set up a report address. Without rua, receivers send no aggregate reports (§4.7). The reports are meant to be machine-read, so you need a report-processing service or your own tooling (§5.1.5). If the address is on another domain, that domain must declare in DNS that it accepts your reports (RFC 9990).
  2. Publish with p=none. The point is to spot, in the reports, a server missed in the initial SPF and DKIM setup or a third-party sender you did not know about (§5.1.5).
  3. Align every legitimate source. Every legitimate stream in the reports should produce at least one, preferably two, aligned identifiers: a DKIM signature with your own domain (d=) and/or an aligned SPF domain (§5.1). RFC 9989 makes this a MUST before enforcement (§5.1.6).
  4. Test quarantine: p=quarantine; pct=0; t=y. t=y has the policy applied one level lower, so the receiver effectively applies none; reports are unaffected (§4.7). pct=0 does the same job for receivers working to RFC 7489: they do not recognise t and would enforce a record carrying only t=y in full.
  5. Move to quarantine: remove pct=0 and t=y together. Mail failing authentication is treated as suspicious; depending on the receiver it goes to spam or quarantine.
  6. Test reject: p=reject; pct=0; t=y. The receiver effectively applies quarantine.
  7. Move to reject: remove pct=0 and t=y together. Mail impersonating your domain is rejected.

The records along the way, in order:

"v=DMARC1; p=none; rua=mailto:dmarc@ornek.com.tr"
"v=DMARC1; p=quarantine; pct=0; t=y; rua=mailto:dmarc@ornek.com.tr"
"v=DMARC1; p=quarantine; rua=mailto:dmarc@ornek.com.tr"
"v=DMARC1; p=reject; pct=0; t=y; rua=mailto:dmarc@ornek.com.tr"
"v=DMARC1; p=reject; rua=mailto:dmarc@ornek.com.tr"

How long it takes

RFC 9989 gives no fixed duration; it ties it to your sending pattern and says that becoming sure all your mail is authenticated “may take many months” (§5.1.7). The criterion is the reports, not the calendar: move on when no unrecognised legitimate source remains. Systems that send monthly or quarterly (invoicing, payroll, period-end notices) are easy to miss for exactly this reason.

Subdomains

p also applies to subdomains unless sp is written. Writing np=reject for non-existent subdomains does not affect your legitimate mail, since no legitimate mail comes from a domain that does not exist; you can add it even while your main policy is still none.

Frequently asked

Can I not phase in with pct?

RFC 9989 has no pct; t replaced it. In practice pct only behaved consistently at 0 and 100 (Appendix A.6), so a percentage rollout with values such as pct=50 is not reliable. During the transition pct=0 still has a use: receivers working to RFC 7489 do not recognise t, so the test steps carry pct=0; t=y together.

Will mailing lists be affected once I move to reject?

They can be: lists that modify messages can break the DKIM signature. RFC 9989 discusses these interoperability issues separately and asks that they be understood before deciding (§5.1.7, §7.4).

Last updated: