Today, a lot of legitimate mail is impossible to tell apart from abuse.
The goal: move legitimate mail out of the grey area — so anything still there can be treated as bad.
Good means: from who it says it’s from, with every change attributable to whoever made it.
Forwarding, mailing lists, Bcc and bounces are legitimate, but today
they look just like the tricks abusers use: replayed signatures, forged
bounces, spoofed From addresses. Receivers have to guess,
and both good and bad mail get caught. DKIM2 gives each legitimate use
a clear, verifiable way to do what it does — forwarders sign,
lists record their changes, each Bcc copy is signed for the recipient
it was sent to, bounces follow the chain — so it leaves the grey
area.
Four things you can’t verify today
- Who the message was for. You can receive legitimate mail with no proof you were the intended recipient — that’s how Bcc works. So a validly signed message can be captured and replayed to huge numbers of other people, riding on the signer’s good reputation.
- Whether a bounce is safe to send. Once the SMTP connection has closed you can’t tell a real sender from a forged one, so bounces get scattered onto innocent parties (“backscatter”).
- What changed in transit. When a forwarder or list
edits a message, the signature simply breaks; you can’t tell
new content from the original. So mailing lists rewrite the
Frominto their own domain and start again. - Whether you can move to new crypto. Nobody can rely on a new signing algorithm until everyone verifies it, so there’s no safe path to stronger, or post-quantum, algorithms.
The first two share a root cause: the SMTP envelope (MAIL
FROM and RCPT TO) is never authenticated.
DKIM2 fixes all four
- The envelope is signed into the message — replay is detectable, and bounces can safely come home.
- Every change is recorded — each part of the message can be reliably attributed to whoever wrote or changed it.
- Signatures carry several algorithms at once — upgrade crypto with no flag day.
Not trust — accountability. None of this makes a message trustworthy. DKIM2 doesn’t tell you who to trust; it records who did what at every step, so reputation lands on whoever actually did it. The rationale is set out in the DKIM2 motivation draft (background; now expired).
Where the work is
DKIM2 is being developed in the re-chartered IETF DKIM working group. It is an active Internet-Draft, not yet a finished standard — details may still change. We expect it to become what the large mailbox providers require.
| Working-group draft | Revision | What it is |
|---|---|---|
| dkim2-spec | -09 | The core specification |
| dkim2-bcp | -01 | Best practices |
| dkim2-dns | -03 | Key records in the DNS |
| Related individual draft | Revision | What it is |
|---|---|---|
| dkim2-authres | -02 | DKIM2 results in Authentication-Results |
| dkim2-debug-header | -03 | X-DKIM2-Info diagnostics for interop |
| dkim2-sender-policy | -04 | DMARC policy for DKIM2 |
It is already running on real mail: see who has deployed DKIM2, and the independent implementations that interoperate.
Get involved
- Join the IETF DKIM working group’s mailing list — this is where it’s decided.
- The feedback-report format is still open: bring your use cases and help shape it.
- Running code & interop — signers, verifiers and shared test vectors.
Learn more
- How DKIM2 works — the chain, recorded changes, the signed envelope
- Get ready — what senders should do now
- Try it — addresses that reflect, originate and bounce DKIM2 mail
- Verify a message in your browser — paste an email; the message never leaves your machine
- Verify a message on our server — the same check, with a per-level breakdown
- Run your own DKIM2 list host — install guide for Postfix with Mailman 3 or Sympa
- DKIM2 for stock Mailman — chosen lists of an unpatched Mailman 3, with the milters and Postfix routing only