Email authentication is rarely broken by an attacker. It is broken by a colleague. Someone in marketing signs up for a webinar platform. A contractor tidies the DNS zone and deletes a record with a strange name. A vendor moves its mail servers and announces the change in a newsletter nobody opened. Each act is innocent. Each one quietly unpicks a setup that took weeks to get right.
Setting up SPF, DKIM, and DMARC is a project, and projects end. What keeps them working is a short list of habits. These are the ones that separate the domains that stay protected from the domains that were protected once, in the spring of some earlier year.
1. Keep a written inventory of everything that sends as you
You cannot authenticate a sender you do not know about. Keep a list. It needs five columns: the service, what it sends, the business owner, how it authenticates (DKIM selector, return-path), and the date someone last confirmed it.
The list will be wrong within a quarter, which is why you check it against evidence. DMARC aggregate reports record who actually sent mail using your domain, as seen by participating receivers. Compare the two regularly. A source in the reports that is missing from the list is either a system someone forgot to mention or someone pretending to be you, and you want to know which. SpoofKiller’s DMARC monitoring keeps that comparison running; the sending sources documentation explains how each source is identified and classified.
Make one rule and enforce it: no new tool sends mail from the company domain until it is on the list and aligned. It is a two-line policy that prevents most future incidents.
2. Treat SPF as a scarce resource
An SPF record gets ten DNS lookups and one record per name. That is a budget, and every vendor’s setup guide asks for a piece of it. Spend it deliberately.
- One owner. A named person approves every change to the record. Records edited by committee end with two
v=spf1lines and a permanent error. - Remove what you cancel. When a contract ends, its include goes the same week. A stale include wastes lookups and keeps authorizing servers you no longer pay for.
- Prefer a custom return-path to an include. If a platform can use a bounce subdomain of yours, its SPF lives there and costs your root record nothing.
- End with
~allor-all. With DMARC enforcing, either is sound. Never?all, and never+all, which authorizes the entire internet. - Skip
ptr, and question everyaandmx. Most were added by reflex.
Check the count after every change with the SPF checker, and rebuild cleanly with the SPF generator when the record has become archaeology. Nested includes change without notice, so a record can exceed the limit while you sleep. Scheduled SPF monitoring catches that; so does a calendar reminder, if you are the kind of person who obeys them.
3. Sign everything with DKIM, using your own domain
SPF fails whenever mail is forwarded. DKIM does not. That makes an aligned DKIM signature the most dependable way to pass DMARC, and the practice follows: every sender signs, and every signature uses your domain in the d= tag.
- Use 2048-bit keys. 1024 is the minimum receivers accept, not a target.
- Give each service its own selector and key. When you drop a vendor, you revoke one key and nothing else is disturbed. A shared key means a shared fate.
- Rotate. Twice a year is a reasonable rhythm for keys you manage. Where a provider rotates through delegated CNAMEs, let it.
- Rotate in the right order. Publish the new key, switch signing, wait about a week for mail in transit, then revoke the old one by emptying its
p=tag. - Sign after the last modification. If a gateway adds a disclaimer, the gateway signs. A footer added after signing breaks the signature every time.
Keep a record of your selectors. They cannot be listed from DNS, so a selector nobody wrote down is a selector nobody can audit. The DKIM checker verifies any one you name, and Common DKIM Errors covers the failures.
4. Put different kinds of mail on different subdomains
Mail from people, mail from applications, and mail from marketing have different risks. Give each a home:
example.com people: staff mailboxes
notify.example.com applications: receipts, password resets
news.example.com marketing: newsletters, promotionsThis arrangement pays three ways:
- Each subdomain has its own SPF record and its own ten lookups, which ends the budget problem.
- Reputation is partly separated. A campaign that draws complaints does less damage to the domain your invoices come from.
- Enforcement can proceed unevenly. A subdomain may publish its own DMARC record, so a troublesome stream need not hold back the rest.
It also contains the damage when a vendor is compromised or simply careless. A vendor authorized on news.example.com cannot send as your chief financial officer.
5. Reach enforcement, then keep reading the reports
A policy of p=none is a measuring instrument. It stops no forged mail. The destination is p=reject, reached in stages; the staged plan lays out the order and the checks.
Along the way:
- Always publish a
ruaaddress, and send reports to a dedicated mailbox or a monitoring service, never to a person’s inbox. - Leave alignment relaxed unless you delegate subdomains to parties you do not trust. Strict mode breaks more legitimate mail than it stops forged mail. See DMARC Alignment Explained.
- Do not read every failure as an attack. Forwarding and mailing lists fail DMARC routinely. Look for change, not for zero.
- Keep watching after reject. Before enforcement, a broken record costs you nothing visible. After it, the same mistake bounces payroll.
Before enforcement, mistakes are free and so is forgery. After it, both get expensive. That is the trade, and it is a good one.
6. Lock the domains that send nothing
Your main domain is well defended. The old brand name, the misspelling you registered defensively, and the country-code variants are not, and a forger is happy to use any of them. A domain that sends no mail can be locked completely in five minutes:
idle-example.com. TXT "v=spf1 -all"
_dmarc.idle-example.com. TXT "v=DMARC1; p=reject; rua=mailto:dmarc@example.com"
idle-example.com. MX 0 .No authorized senders, a reject policy, and a null MX saying the domain takes no mail. Do the same for subdomains that should never send, or cover them all at once with sp=reject on the parent. Run the full set through the domain health check once a year; idle domains are where forgotten records collect.
7. Guard the DNS underneath it all
Every control above is a DNS record. Whoever can edit your zone can rewrite your SPF, replace your DKIM keys, and set your DMARC policy to none. Protect the zone accordingly:
- Limit who can edit DNS, require multi-factor authentication at the registrar and DNS host, and turn on registrar lock.
- Review changes. A second pair of eyes on a TXT record edit is cheap. So is a note saying what each odd-looking record is for, which stops the well-meaning cleanup.
- Lower the TTL before a planned change, and raise it afterward, so a mistake can be reversed in minutes.
- Enable DNSSEC where your registrar and host support it, so answers cannot be forged in transit.
- Renew early. An expired domain takes all its mail with it and may be bought by someone else.
The DNS security checker reports DNSSEC, CAA, and nameserver status. SpoofKiller’s domain security monitoring re-checks the records your mail depends on every six hours, keeps a history of changes, and warns of approaching expiration.
8. Add the layers beyond DMARC
With the foundation in place, three additions are worth the effort:
| Standard | What it does | When to add it |
|---|---|---|
| MTA-STS | Tells sending servers to require encrypted, verified connections when delivering mail to you | Once inbound mail is stable; begin in testing mode |
| TLS-RPT | Asks senders to report failed encrypted deliveries to you | Alongside MTA-STS, or before it |
| BIMI | Displays your logo beside authenticated mail in supporting inboxes | After DMARC is at quarantine or reject |
MTA-STS and TLS-RPT protect mail on its way to you. DMARC protects your name on mail claiming to come from you. They solve different problems and do not substitute for one another. BIMI is the visible reward for the invisible work; the BIMI checker tells you whether your policy qualifies.
None of this makes a domain immune. Authentication stops exact-domain forgery. It does not stop lookalike domains or stolen passwords, which are covered in Domain Impersonation Protection. SpoofKiller’s Protection Score summarizes how much of this list a domain has in place on a scale of 0 to 100. A perfect score means the list is done. It does not mean you can stop reading the reports.

