Skip to content
SpoofKiller
Email spoofing detection

Not every failed message is a forgery. We show you which ones probably are.

A tool that cries spoofing at every DMARC failure is wrong much of the time and ignored all of the time. SpoofKiller weighs the evidence for impersonation against the evidence for forwarding and plain misconfiguration, and shows its work.

Shows whether your current DMARC policy lets receivers refuse forged mail.

example.com / findings
HighConfidence 80%

Potential Domain Spoofing Detected

203.0.113.45 · provider not identified

Why it was flagged

  • 318 messages failed both SPF and DKIM alignment
  • The source had not been seen before
  • It matches no recognized email provider
  • No DKIM signature was present at all

Reasons for doubt

  • No forwarding override was reported
  • Seen by one reporting organization so far
Mark ApprovedUnder InvestigationMark Unauthorized
Illustrative example: sample values, not customer data.
The problem

Three innocent explanations and one guilty one, all wearing the same coat.

Forging the From address of an email requires no skill. The protocol permits it. Anyone with a mail server can put your domain there, and unless you have published an enforced DMARC policy, many receivers will deliver the result. That is the guilty explanation for a failing message.

The innocent ones are everywhere. A customer forwards your invoice from one mailbox to another, and SPF fails because the forwarder is not in your record. A mailing list adds a footer and breaks the DKIM signature. Your own marketing platform sends on your behalf with nobody having finished its setup. In a raw DMARC report, all four look identical: a row, an IP address, and the word fail.

Treat them all as attacks and you will spend your week chasing a university's forwarding server. Treat them all as noise and you will miss the one that mattered. The useful work is telling them apart, and preventing spoofing begins with knowing which failures to fix and which to refuse.

A halftone paper envelope wearing a black carnival mask on a red background, standing in for forged email.
The mask is cheap. The envelope underneath is where the evidence is.
What SpoofKiller does

A scoring rule you can read in full.

Every source with mail that failed both SPF and DKIM alignment starts at 35. Evidence moves the number up or down. The result is held between 5 and 95, because certainty is not on offer.

Raises confidence

  • The source had not been seen before+15
  • It sends from a network new to this domain+10
  • It matches no recognized email provider+10
  • No DKIM signature was present at all+10
  • SPF authenticated an unrelated domain+5
  • 100 or more failing messages+10
  • You previously marked it Unauthorized+25

Lowers confidence

  • The receiver noted a forwarding or mailing-list override-30
  • A DKIM signature for your domain was present but broken-20
  • The source is a large mailbox provider-15
  • It confidently matches a known sending service-10
  • Fewer than 5 failing messages-10
Below 40

Not flagged as impersonation.

40 to 59

Flagged: unauthenticated mail from an unfamiliar source.

60 and up

Flagged: Potential Domain Spoofing Detected.

Sources you have marked Approved or Ignored are never flagged by this rule. Severity rises to high only when confidence is 75 or more and at least 50 messages failed. Names for sources come from sender identification, which carries its own confidence figure.

Four more rules

The failures that are your own fault get their own label.

An approved service that fails DMARC is not an intruder. It is a to-do item, and it is the one that will hurt you when you tighten the policy.

A radar sweep behind a shield.
  • Configuration, not attack

    Approved sender failing authentication

    A service you approved, or one confidently identified as a known sending platform, fails DMARC on more than a fifth of at least 20 messages. This is your own tool, half set up. The finding says so and tells you to configure a custom DKIM domain or return-path.

  • Informational

    Newly observed sending source

    A source sends as your domain for the first time and does not trip the impersonation rule. Raised only after the domain has a week of history, so onboarding does not produce a flood.

  • Worth confirming

    Signing domain changed

    A source seen for more than two weeks begins signing with a DKIM domain it has never used. Usually a provider change. Occasionally not.

  • Trend

    Authentication failure spike

    A day's failures reach at least 50 messages and at least three times the recent daily average, measured against a trailing 14-day baseline with five or more days of data.

How it works

Report in, verdict out, your decision remembered.

  1. 01

    A report arrives

    Each aggregate report is parsed and its rows are matched to sending sources. Sources never seen before are marked new.

  2. 02

    Each failing source is weighed

    For every source with mail that failed both SPF and DKIM alignment, the rules add points for suspicious evidence and subtract points for innocent explanations.

  3. 03

    A finding is raised, once

    At a confidence of 40 or more, a finding is created with both lists of evidence. The same source seen again updates that finding instead of creating another.

  4. 04

    You classify, and it learns your answer

    Mark the source Approved, Under Investigation, Unauthorized, or Ignored. Approved and Ignored sources stop being flagged; Unauthorized ones score higher next time.

What you see for each finding is in the example at the top of this page: a title, a severity, a confidence figure, the reasons it was flagged, and the reasons for doubt. Beneath that sit the source's IP address, provider, network, first and last seen dates, message counts, and the organization that reported it. Findings feed alerts, which are deduplicated so one persistent source does not become forty emails. All of it rests on the report data from DMARC monitoring.

Honest limits

What the rules cannot know.

Detection works from counts and authentication results. That is a good deal, and it is not everything.

  • It cannot prove intent

    A high score means the evidence fits impersonation better than the alternatives. It is a reason to look, not a conviction. Findings are marked as uncertain for that reason.

  • It does not block mail

    Receivers enforce; you set the policy. Getting from none to reject is the actual defense, and Enforcement Readiness tells you when the step is safe.

  • Lookalike domains are out of scope here

    Mail from a near-copy of your domain never appears in your DMARC reports. That is handled by lookalike discovery, and explained in Domain Impersonation Protection.

  • Compromised accounts pass

    Mail sent by an attacker from a real, hijacked mailbox at your provider is authenticated mail. DMARC, and therefore this feature, will see nothing wrong with it.

  • It sees only reported mail

    A forgery campaign aimed entirely at receivers that send no aggregate reports leaves no trace in the data.

  • It runs a day behind

    Aggregate reports are typically sent every 24 hours. Detection happens when the report is processed, not when the message is sent.

Try it before you buy it

First question: can your domain be forged at all?

The free domain health check reads your DMARC, SPF, and DKIM records and tells you whether receivers have been asked to refuse mail that fails. If the policy is none, the answer is that they have not.

Free. No account. The lookup runs against live DNS.

Free, no obligation

Seen a forged email from your own domain?

Tell us the domain and what happened. We will check what your records currently allow and write back with the shortest route to closing it.

  • A person reads every message
  • No sales sequence, no list rental
  • Reply within one business day

We use your details to reply. Nothing else.

FAQ

Questions about spoofing detection

Straight answers. If yours is not here, ask us and a person will reply.

Ask a question

Find out who is sending mail in your name.

The check is free and takes about as long as reading this sentence. No account, no card, no sales call.