Homoglyph domains and IDN homograph attacks
A homoglyph domain replaces letters with characters that render identically or nearly so: rn for m, 0 for o, or a whole label spelled in Cyrillic letters that mirror Latin ones. No typo is ever typed; the swap is built into the characters themselves.
Updated
- Encoding tell
- Punycode labels start with xn--
- Classic pair
- Cyrillic о (U+043E) vs Latin o (U+006F)
- Tracked as
- MITRE ATT&CK T1583.001
Most lookalike techniques in typosquatting assume someone mistypes. Homoglyph domains assume someone misreads. The registrant swaps characters for lookalikes drawn from the same alphabet or another script entirely, producing a domain that renders the same as yours to anyone reading it.
How the substitution works
Unicode assigns characters to many scripts, and many of them look alike. Latin o (U+006F) and Cyrillic о (U+043E) are different characters that Unicode’s confusables data lists as visually confusable, and the same data maps Greek omicron and Armenian oh to Latin o. Registries such as Verisign reject a label that mixes Latin and Cyrillic characters, so the workable attacks take two forms:
- Whole-script lookalikes. Every letter comes from one script that happens to mirror Latin. In 2017 Xudong Zheng registered
аррӏе.com, spelled entirely in Cyrillic, and showed that Chrome, Firefox and Opera displayed it in a form indistinguishable fromapple.com. - ASCII confusables. No Unicode at all:
rnform,0foro,1orIforl. These are ordinary registrations under any TLD and render exactly as typed.
Either way, the result is a different registration, owned by someone else, that can look right in an email, an ad, or a browser tab.
Attackers use these domains for phishing and malware delivery, and a visitor cannot tell the site is fraudulent without carefully inspecting its URL or certificate. MITRE ATT&CK catalogs acquiring domains for impersonation under technique T1583.001 (Acquire Infrastructure: Domains), homograph registrations included.
Why Punycode hides it
DNS lookups use ASCII hostnames, so internationalized labels are encoded as Punycode: the all-Cyrillic аррӏе is stored and transmitted as xn--80ak6aa92e. Two consequences follow:
- The raw form betrays the trick. Any label starting
xn--is an internationalized name, and decoding it reveals which characters were swapped. - Readers usually see the decoded form. Browsers display the Unicode form whenever their IDN policy allows it, which is what makes the attack work.
Browsers push back with display policies. Chrome applies the Highly Restrictive profile from Unicode’s security mechanisms standard (UTS #39), so Latin mixed with Cyrillic or Greek shows in its xn-- form. It also shows Punycode when an all-Cyrillic label is built only from Latin lookalike letters and the TLD is neither Cyrillic nor one known to host many Cyrillic domains, and when a label’s confusable skeleton matches a top domain. Other browsers apply their own rules. Display policies only choose between Unicode and Punycode, so ASCII confusables, which contain nothing to decode, pass through them; Chrome’s separate lookalike warning covers hostnames whose skeleton matches a top 500 domain or a site the user recently engaged with.
What registries do, and what is left
ICANN’s IDN Guidelines require registries offering IDNs under ICANN registry agreements to take all code points in a label from one script, with exceptions for languages that mix scripts, and Verisign, the .com registry, applies a single-script rule. What remains are whole-script lookalikes, which a single-script rule cannot stop and which the guidelines only encourage registries to limit, and ccTLDs, for which the guidelines are best current practice rather than a requirement. A registrant needs only one permissive combination of script and TLD, and ASCII confusables need no IDN support at all.
You can enumerate the ASCII confusables of your own domain, such as 0 for o, 1 for l and rn for m, with the lookalike domain generator. In registration monitoring, notolens also maps internationalized registrations to their Latin lookalike form, so an IDN whose letters spell your brand surfaces in your inbox like any other match. Reporting it to the sponsoring registrar follows the same abuse-desk process as any deceptive domain.
Sources
- Unicode Technical Standard #39: Unicode Security Mechanisms, restriction levels and confusables
- Chromium: IDN display policy in Google Chrome
- Verisign: Internationalized Domain Name registration rules
- ICANN: Guidelines for the Implementation of Internationalized Domain Names, version 4.1
- Xudong Zheng: Phishing with Unicode Domains (2017)
- RFC 3492: Punycode encoding for internationalized domain names
- MITRE ATT&CK T1583.001: Acquire Infrastructure: Domains
- WIPO Overview 3.1: panel views on UDRP questions, section 1.9
- Unicode: confusables.txt, the confusable mappings data for UTS #39
- IANA: Delegation record for .com
Frequently asked questions
A homoglyph is a single character that looks like another; a homograph is a whole string that looks like another string. An IDN homograph attack registers a domain built from homoglyphs so the rendered name is indistinguishable from the target. The terms overlap in practice and the defenses are the same.
Partially. Chrome and other browsers apply IDN display rules that render a label as Punycode (its xn-- form) when it mixes scripts like Latin and Cyrillic, or when an all-Cyrillic label mimics Latin letters under a TLD not associated with that script. The rules differ by browser and only decide how internationalized labels are displayed, so they do not cover ASCII lookalikes such as rn for m, and visual inspection is not a reliable defense.
Not practically. A label with several confusable positions produces thousands of combinations, multiplied again across TLDs. Detection and takedown are more tractable than defensive registration at that scale.
Yes. The WIPO Overview lists the use of non-Latin internationalized or accented characters, and of similar-appearing characters such as numbers used to look like letters, among the variations of a mark that panels consider confusingly similar under the first UDRP element.
Related tools and resources
Lookalike Domain Generator
Enumerate homoglyph and typo variants of your own domain, in your browser.
Typosquatting
The full range of lookalike registration techniques, scale data, and defenses.
How to Report Domain Abuse to a Registrar
Find the sponsoring registrar and submit an actionable abuse report.
Ongoing brand monitoring
notolens checks new domain registrations across 1,570 TLDs and app store results daily, search ads multiple times a day, and trademark registers weekly. When a lookalike domain, conflicting mark, copycat app, or ad using your name appears, notolens checks it, explains the risk, and hands you the records and possible next steps.