The Email Parser Nightmare: Unicode, RFC Quirks and Account Takeover
A single email address can hand over an account. The trick is that it passes through three systems that each normalize it differently. You wouldn't think an email address could trigger a zero-click account takeover. It'
A single email address can hand over an account. The trick is that it passes through three systems that each normalize it differently.
You wouldn't think an email address could trigger a zero-click account takeover. It's just a string: [email protected]. But that string travels through at least three systems:
- the frontend validator (regex / HTML5),
- the auth and database layer (SQL, Mongo, your OAuth provider),
- the delivery engine (SES, SendGrid, Postmark, raw SMTP).
When those three normalize differently, authentication collapses. Here are the edge cases worth testing.
1. IDN homograph and Punycode collisions
Register with а[email protected] — the first letter is U+0430, Cyrillic "а", not ASCII "a".
- If the system stores it literally and never canonicalizes to Punycode (
xn--dmin-43d), it passes uniqueness checks as a brand-new account. - But if some casing or folding step folds that Unicode character improperly (Turkish dotless i, Cyrillic folding), it can normalize to the real admin's identifier.
2. Zero-width and whitespace steganography
victim@example[u200B].com zero-width space
victim[u00A0]@example.com non-breaking space
victim[x00]@example.com null-byte truncation
The auth database strips the unprintable character and sees [email protected]. The mailer keeps it and delivers the reset link somewhere else.
3. Gmail dot-blindness
Gmail ignores dots: [email protected] is the same inbox as [email protected]. If an app treats them as two users but allows Google SSO, an attacker can register the dotted variant first and later hijack pre-seeded state, invites or payment methods when the victim signs in with OAuth.
4. RFC 822 header smuggling
Two classics:
-
CRLF injection:
[email protected]%0d%0aBcc:[email protected]. Concatenated straight into SMTP headers, that injects aBcc:and silently copies every reset email to the attacker. - Display-name abuse: some lax parsers accept a display name plus a bracketed mailbox, then a second mailbox after it, and deliver to the wrong one.
How to fix email normalization
- Canonicalize IDNs to ASCII Punycode at the boundary, before any database lookup.
-
Strip invisible and control characters — zero-width (
u200B,u200C,u200D), non-breaking space (u00A0), control codes (x00–x1F). - Case-fold in ASCII, not with locale-dependent Unicode folding.
- Never interpolate addresses into raw mail templates. Use a library that parameterizes headers.
Ever caught an auth bypass or reset hijack from an email-parser mismatch? Share it below.
Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.