TL;DR: SPF hardfail, DMARC reject, TLS reporting and CAA are four lines of DNS that close the cheapest half of phishing. They do not stop a mail sent from the real domain, and they do not stop a display name. Knowing which half you closed is the difference between protection and the feeling of it.
The scenario
I changed four records on a domain I control, then sent mail through the normal path to see what the receiver would say. The whole change took ten minutes and cost nothing.
The interesting part was not what improved. It was drawing the line afterwards: which attacks are now impossible, and which are exactly as easy as before.
The four records
spf "v=spf1 include:provider-a include:provider-b -all"
dmarc "v=DMARC1; p=reject; sp=reject; adkim=r; aspf=r;
fo=1; rua=mailto:dmarc@example.tld; pct=100"
tls-rpt _smtp._tls "v=TLSRPTv1; rua=mailto:dmarc@example.tld"
caa 0 iodef "mailto:dmarc@example.tld"
Each one answers a different question:
- SPF
-all— may this envelope domain send? A hard fail refuses instead of marking. - DMARC
p=reject— may this visible domain send, and what happens if it cannot prove it? Reject removes the message instead of filing it. - TLS-RPT — did the transport use TLS, and did it fail anywhere? Reporting only, but it makes silent downgrades visible.
- CAA — which certificate authorities may issue for this domain? It limits mis-issuance and records who to notify.
Prove it before you celebrate
The only useful proof is a real message through the real path after the change:
labels: INBOX, CATEGORY_UPDATES
dkim=pass header.i=@example.tld header.s=esp
spf=pass ... @send.example.tld ...
dmarc=pass (p=REJECT sp=REJECT dis=NONE)
The receiver confirms the new policy and still delivers. That single line — p=REJECT in the authentication result — is the difference between "we published a policy" and "the other side is enforcing it".
What the four records close
| Attack | Before | After |
|---|---|---|
| forged sender, direct to the MX | 550 | 550 (unchanged) |
| forged sender through a rented relay | delivered to spam | rejected at SMTP |
| envelope spoofing | marked as softfail | refused |
| certificate mis-issuance | unconstrained | limited by CAA |
| silent TLS downgrade | invisible | reported |
The relay case is the one that actually changed. Direct spoofing from an arbitrary IP was already refused by the large receivers before DMARC was even evaluated — PTR and authorization checks come first.
What they cannot fix
| Attack | Why DNS cannot help |
|---|---|
| mail from the real mailbox | the message is genuinely signed |
| brand in the display name | no record covers free text |
| lookalike domain | it belongs to someone else |
| a user who clicks | not a name-resolution problem |
| a compromised ESP key | it produces valid signatures |
That second row is the one people underestimate. DMARC checks the domain, not the name in front of it. A correctly authenticated domain can wear any display name it likes.
Findings
| Layer | Status after the change |
|---|---|
| forgery from outside | closed |
| drift | monitored (reports + posture check) |
| impersonation with your name | open |
| impersonation with your mailbox | open |
| transport visibility | improved, not enforced |
The chain
The parts that need more than DNS
Four items always come up in the same conversation, and none of them is a DNS record:
- DNSSEC needs a registrar that can publish a DS record. If the registrar cannot, the zone stays unsigned no matter what the DNS provider supports.
- MTA-STS needs a policy file on a subdomain, which means a host — a web server or a small worker.
- BIMI needs a verified mark certificate, which is a commercial process.
- Two-factor enforcement needs a click in an admin console. It is also the only control that touches the mailbox-takeover path.
Closing
Four records, ten minutes, zero cost — and a measurable improvement in one specific class of attack.
That is worth doing. It is not worth confusing with "we are protected". The honest scorecard has two open rows, and both of them live where DNS cannot reach: in the login and in the name.
Once an attacker gains access, they can perform the following actions: