ZeroName

How it works

ZeroName is an attribution tool. It does not keep anything secret. It marks a name with a hidden tag so you can identify which recipient leaked your contact details. This section describes what the encoder and decoder do.

The invisible codepoints

The payload is stored in two ranges of zero-width characters. They display nothing but survive copy and paste as normal text:

• U+FE00 … U+FE0F, Variation Selectors 1–16 (VS1–VS16), hold byte values 0–15.
• U+E0100 … U+E01EF, Variation Selectors Supplement, hold byte values 16–255.

Variation selectors normally select a glyph variant of the character before them. When there is no such variant, they display as nothing. They are also normalization-stable: they have no compatibility decomposition, so NFC and NFD leave them unchanged. NFKC does remove them, which is covered in the caveats.

Encoding

1. UTF-8 encode the domain into bytes.
2. Compute a checksum byte by XORing every domain byte together, and append it as the final payload byte.
3. Derive a keystream from SHA-256(passphrase) using crypto.subtle.digest, then repeat those 32 bytes as needed to cover the payload.
4. XOR each payload byte (domain bytes and checksum) with the matching keystream byte.
5. Map each resulting byte to one invisible codepoint: 0–15 to U+FE00 + b, and 16–255 to U+E0100 + (b − 16).

This produces one invisible codepoint per payload byte. A 12-byte domain plus its checksum becomes 13 hidden characters.

Interleaving between the letters

The hidden characters are placed in the gaps between the visible letters, never before the first letter and never after the last. They are spread evenly across those gaps, so hidden character k is attached to the letter at floor(k × gaps ÷ payload). Encoding the domain linkedin into the name Noah places three invisible bytes in each gap, giving N•••o•••a•••h, where each • is an invisible byte attached to the letter before it.

The placement is monotonic, so the hidden characters stay in payload order. Decoding does not depend on where they sit: it reads the string by codepoint, collects the characters in the two invisible ranges in order, and that sequence is the payload. There is a single copy of the payload, so if any hidden character is lost, the checksum fails.

A name needs at least two visible characters to have a gap. If it has only one, the payload is appended after it.

The keyed XOR keystream

SHA-256 turns the passphrase into 32 bytes. Those bytes are cycled across the payload and XORed with it, and the same passphrase produces the same keystream to reverse the operation. Without the passphrase the bytes do not decode, and a tag made with a different passphrase will not decode against yours. This is obfuscation rather than encryption. A repeating-keystream XOR can be broken by anyone who knows the scheme and has enough samples. It stops the domain from being read directly, but it does not resist analysis.

The checksum

The last payload byte is a checksum. On decode, the XOR of the recovered domain bytes is compared against it. If a character was added, a byte was changed, or the passphrase is wrong, the values differ and the decoder reports Invalid or stripped tag instead of showing an incorrect domain. It is a one-byte check and does not correct errors.