How Does Invoice Fraud Use Lookalike Domains to Divert Payments?
Published 2026-11-14 · Reviewed 2026-11-14 · 6 min read · By the Ingenious AI team
Is someone impersonating your domain right now?
Run the free DomainGuard scan: lookalike registrations, MX-armed variants, certificate-transparency and CAA checks for your domain, in one report.
Run my free DomainGuard scanThe short answer
Invoice fraud diverts payments by inserting one false message into a real money conversation. In the lookalike variant, a supplier-shaped domain — one character different — emails your accounts team new bank details just before a genuine invoice falls due. In the thread-hijack variant, a compromised mailbox continues a real thread. Both end the same way: your payment leaves for the attacker's account on your own instruction. The control that breaks both is verifying bank details by callback on a number you already hold, applied to every change, every time, without exception for urgency or seniority.
The crime scene is a thread your team trusts
Unlike most phishing, invoice fraud does not need to create belief from nothing. The payment is real, the amount is expected, the supplier relationship is genuine — the attack changes exactly one variable, the destination account, inside a context your processes already approved. That is why amounts are large: the fraud rides transactions that were already budgeted, from building progress payments and equipment purchases to the professional-services invoices every company pays monthly.
Timing is researched, not lucky. Attackers learn due dates from compromised mailboxes, from documents businesses share carelessly, or simply from the rhythm of monthly billing — a 'bank change' notice arriving two days before your usual payment run needs no insider knowledge at all. The message itself is short, clerical, and boring, which is the point: bank-detail notifications are supposed to be administrative, and administrative messages get processed, not interrogated.
Two variants, one destination
The lookalike variant needs no compromise of anyone's systems. The attacker registers the supplier's near-name, sets up MX and authentication so the mail is technically impeccable, and writes to your payables desk as the supplier's accounts contact. Everything about the message is fake except the invoice it references — which may be copied from a real one obtained through any earlier leak, or simply invented plausibly against a known contract.
The hijack variant starts inside a real mailbox — yours or the supplier's — and is worse: the fraudulent instruction arrives within the genuine thread, from the genuine address, sometimes replying to your own last question. No domain defence catches it, because nothing about the domain is wrong. This variant is why invoice-fraud defense can never be purely technical, and why the verification control must be indifferent to how authentic the message looks, sounds, or threads.
Why authentication does not save you
It is worth stating plainly, because smart teams get this wrong: a lookalike's mail passes SPF, DKIM, and DMARC. Those protocols verify that a message is authorised by the domain it claims — and the attacker owns the claiming domain. Your DMARC enforcement protects your name from being spoofed exactly; it does nothing about your supplier being approximated. Gateway lookalike-flagging helps, banner systems help, monitoring your suppliers' name-space is beyond your control entirely. The residual risk lands, inevitably, on process.
That is not a counsel of despair; it is a targeting instruction. Spend the technical budget on making the fraud visible — banners, near-match flags on supplier domains, monitoring of your own lookalikes for the outbound mirror of this crime — and spend the process budget on the one control with a perfect record when actually applied: independent verification of bank details, on contact information that predates the request.
The callback control, engineered properly
Weak version: 'we usually ring them.' Strong version: bank details for any payee are entered or changed only against a verified source — a call to the number on the supplier's original contract or previous invoice, answered by someone who can confirm, with the confirmation logged against the payment. New payees get the same treatment at setup. Changes requested by email, letter, or portal message all route through the identical gate, because the fraud happily arrives by any channel that bypasses the habit.
Engineer out the failure modes. The number in the fraudulent message is never the verification number. A callback that reaches voicemail is not a verification. Urgency shortens nothing — the control exists for urgent requests most of all, since urgency is the fraud's favourite lever. And the two-person rule above a threshold ensures that even a successful deception of one clerk meets a second human whose job is to ask for the verification log before releasing the payment.
After a diversion: hours decide recovery
If a payment has gone, the response clock is brutal and simple. Call your bank's fraud line immediately — recalls work when the funds are still moving or freshly landed, and interbank recall success falls steeply after the first 24 to 48 hours. Preserve the fraudulent thread and headers untouched, notify your insurer the same day, and file with law enforcement; recovery and claims both run on documentation created now, not reconstructed later. Tell the real supplier — they may be mid-campaign with their other customers as the unwitting costume.
Then run the blameless debrief that fixes the system: which variant was it, which layer should have caught it, what will be different on Monday. Companies that treat the clerk as the failure get concealment next time; companies that treat the process as the failure get reports. In the cases DomainGuard sees from the monitoring side, the saddest pattern is the firm that discovers, during the debrief, that the lookalike had published MX records three weeks earlier — visible, in public DNS, the entire time.
Five actions to take today
- Require callback verification on pre-existing contact details for every bank-detail creation or change, and log the confirmation against the payment.
- Never verify using a phone number supplied in the requesting message, and treat voicemail as no verification.
- Apply a two-person release rule above your payment threshold, with the second person checking the verification log.
- Turn on gateway flagging for sender domains that near-match your key suppliers, not just your own brand.
- Rehearse the first-hour response to a completed diversion: bank recall, evidence preservation, insurer, law enforcement.
Worked example: the $86,400 invoice that a callback killed
A fictional construction supplier — call it Redrock Building Co, a customer of a firm monitoring 8 domains — appeared to email new bank details ahead of an $86,400 materials invoice. The sending domain differed by one letter; its mail passed every check. The payables clerk followed the standing rule and rang the number on the original contract: Redrock had changed nothing. The genuine invoice was paid to the genuine account two days later. The firm also learned, from the lookalike's DNS history, that its MX records had appeared 19 days before the attempt — a warning that gateway flagging has since been configured to surface automatically for its top twenty suppliers.
FAQ
How do attackers know our payment amounts and dates?
From compromised mailboxes, leaked documents, predictable monthly cycles, and sometimes pure plausibility against a known contract. Assume the context will look right and design controls that do not depend on spotting errors in it.
Does paying by cheque or portal avoid this?
Portals help if access is controlled and bank details inside them are change-logged and verified. The fraud adapts to any channel where a detail can be altered on an unverified instruction — the gate must sit on the change, not the channel.
Our supplier's mailbox was the compromised one. Is it our loss?
Legally it varies by contract and jurisdiction, which is why insurers and lawyers get involved early. Operationally, the lesson is the same: your verification control is the only layer entirely within your power.
Should we verify every invoice, or only bank changes?
Bank changes and new payees are the critical gate — the amount can be perfectly genuine while the destination is false. Verifying destination data once, properly, covers every subsequent invoice to that account.
How do we get staff to comply during busy periods?
Make verification fast, scripted, and praised; make skipping it the only sanctioned behaviour. Fraud peaks exactly when desks are busiest, so the control must be at its most automatic then.