What Are Homoglyph Domains and How Do Attackers Use Them?
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
A homoglyph domain replaces one or more letters in a brand name with characters from another alphabet that look identical on screen — a Cyrillic а instead of a Latin a, for example. Browsers and mail clients display the fake name exactly like the real one, while underneath, the domain is completely different and owned by the attacker. These names are registered through internationalised domain name (IDN) support and travel through DNS in an encoded form starting with xn--, called punycode. Detection means generating the lookalike character substitutions for your own brand and watching those names and their certificates.
The trick: same pixels, different domain
Unicode contains thousands of characters, and many scripts independently developed letters that happen to draw the same shape. Cyrillic а, е, о, р, с, and х are pixel-identical to Latin a, e, o, p, c, and x in most fonts. Greek omicron matches o; Armenian letters supply several more. An attacker who swaps even one of these into your brand name gets a domain that every human reads as yours and every computer treats as a stranger's.
Internationalised domain names make this registrable. DNS itself is ASCII, so the Unicode name is encoded as punycode — the xn-- prefix you sometimes see in certificate logs and WHOIS records. That encoding is your detection friend: legitimate international brands use IDN deliberately and consistently, while a punycode name that decodes to something mimicking an ASCII brand is suspicious almost by definition.
Why browsers do not simply stop this
Modern browsers do try. Chrome and Firefox apply IDN display policies: names mixing scripts in ways typical of spoofing are shown in raw punycode instead of the pretty Unicode form, which gives the game away to anyone who looks at the address bar. But the policies have seams. A name written entirely in one foreign script can still display in Unicode even if it perfectly mimics a Latin brand, because browsers cannot assume every single-script Cyrillic name is an attack — most are legitimate sites for real audiences.
Email is the softer target. Mail clients vary widely in IDN handling, many render the display form prominently and the actual address faintly, and mobile screens truncate the rest. A homoglyph domain sending a payment-change email does not need to fool a careful examiner; it needs to fool a busy person glancing at a phone. That asymmetry — perfect visual copy, hurried human — is why attackers keep registering these names despite browser mitigations.
What attackers build on homoglyph names
The registration is cheap and the uses are the standard impersonation set: credential-harvesting clones of login pages, invoice and payment-diversion email, fake support portals, and malware delivery dressed as document shares. Because the domain renders perfectly, screenshots of the attack often circulate as warnings with comments saying they cannot see anything wrong — which is precisely the point, and precisely why training staff to 'look carefully at the domain' fails against this family.
Homoglyph names also pair with certificates. A domain-validated TLS certificate is automatic for whoever controls the domain, so the fake site gets a padlock and an HTTPS address like any real service. Certificate Transparency logs therefore become a practical early-warning source: a newly issued certificate for a punycode name decoding to your brand's shape is detectable the day it appears, long before the phishing email goes out.
How to detect the variants of your own brand
Detection starts by generating the substitution set: for each character in your brand name, list its cross-script lookalikes, then generate single and double substitutions — the realistic attack space, since names with too many swaps start failing browser display rules and looking odd in punycode. Tools in the typosquatting family, including the permutation engine DomainGuard runs, generate these alongside typo and combosquat variants, then check registration, hosting, and certificates.
When you review results, decode every xn-- name before judging it. A punycode string tells you nothing at a glance; decoded, it either clearly mimics your brand or clearly does not. Prioritise variants that are registered, carry MX records (an email attack being prepared), or hold a fresh certificate. A registered homoglyph with none of those is a watch-list item; one with all three is an incident, and the response is the takedown ladder in our related guides.
Defenses that actually cover this family
You cannot register every Unicode combination of your name — the space is far too large. What works is layered: continuous monitoring of registrations and CT logs for decoded lookalikes, defensive registration of the handful of single-swap variants on major scripts, and mail-flow controls that do not depend on human eyesight. DMARC will not help here, because the homoglyph is a different domain that passes its own checks; gateway rules that flag external mail visually matching your brand, and verification callbacks for payment changes, are the controls that bite.
Finally, tell your customers where you actually live online. A published, stable list of your real domains and sending addresses gives suspicious recipients something to check a homoglyph against. It will not stop the registration — nothing you do can — but it shortens the distance between a customer seeing a fake and a customer reporting a fake, and reports are how takedowns start.
Five actions to take today
- Generate the homoglyph substitution set for your brand across Cyrillic, Greek, and Armenian lookalike characters.
- Decode every xn-- (punycode) result before judging it, and record which variants are registered.
- Flag registered variants with MX records or fresh certificates as priority threats, not curiosities.
- Defensively register the highest-risk single-character swaps you can reasonably hold.
- Monitor Certificate Transparency logs continuously for new certificates issued to lookalike punycode names.
Worked example: Northbeam Outfitters finds a Cyrillic twin
Northbeam Outfitters, a fictional outdoor retailer with 5 domains monitored, ran a full permutation scan that generated 2,300 variants of its brand, including homoglyph substitutions. Eleven variants were registered. One — a single Cyrillic substitution in the brand name — had been registered 9 days earlier, published MX records, and received a TLS certificate 2 days before the scan. Screenshots captured during the takedown filing showed a pixel-perfect clone of the retailer's login page. The registrar suspended the domain in 3 days, before any customer report of phishing mail from it had reached the company.
FAQ
Are homoglyph domains illegal to register?
Registration itself sits in a grey zone, because many IDN names are legitimate. Using one to impersonate and defraud is actionable, and registrars suspend such names on evidence of phishing; trademark dispute processes also apply.
Why does the fake site have a valid padlock?
TLS certificates are issued to whoever controls a domain, automatically and without brand checks for domain-validated certificates. The padlock proves the connection is encrypted to that domain — not that the domain is the brand it resembles.
Can I block these in my email gateway?
Many gateways can flag or rewrite external senders whose domains are visual confusables of your own, and all can banner external mail. Combined with payment-verification callbacks, that covers the main homoglyph email risk.
How do I read a punycode name?
Use any IDN converter: paste the xn-- name and it will show the decoded Unicode form. Certificate Transparency search tools display decoded names alongside the punycode form.
Should I register homoglyph versions of my own brand?
Register the small set of most plausible single-character swaps if budget allows, but treat registration as secondary. Monitoring finds the rest of the space, which is far larger than any portfolio you could buy.