confusable-homoglyphs 1.0

Detect confusable usage of unicode homoglyphs, prevent homograph attacks.

a homoglyph is one of two or more graphemes, characters, or glyphs with shapes that appear identical or very similar wikipedia:Homoglyph

Unicode homoglyphs can be a nuisance on the web. Your most popular client, AlaskaJazz, might be upset to be impersonated by a trickster who deliberately chose the username ΑlaskaJazz.

• AlaskaJazz is single script: only Latin characters.

• ΑlaskaJazz is mixed-script: the first character is a greek letter.

You might also want to avoid people being tricked into entering their password on www.microsоft.com or www.faϲebook.com instead of www.microsoft.com or www.facebook.com. Here is a utility to play with these confusable homoglyphs.

Not all mixed-script strings have to be ruled out though, you could only exclude mixed-script strings containing characters that might be confused with a character from some unicode blocks of your choosing.

• Allo and ρττ are fine: single script.

• Alloτ is fine: mixed script, but τ is not confusable.

• Alloρ is dangerous: mixed script and ρ could be confused with p.

Documentation

confusables

from confusable_homoglyphs import confusables

confusables.is_mixed_script

confusables.is_mixed_script(unicode_string)

Boolean: is unicode_string mixed-script.

confusables.is_confusable

confusables.is_confusable(unicode_string, greedy=False, preferred_aliases=[])

Takes a character or string and returns each character present in unicode’s confusable characters list.

If greedy=False, it will only return the first confusable character found without looking at the rest of the string, greedy=True returns all of them.

preferred_aliases=[] can take an array of unicode block aliases to be considered as your ‘base’ unicode blocks:

• considering paρa,

  • with preferred_aliases=['latin'], the 3rd character ρ would be returned because this greek letter can be confused with latin p.

  • with preferred_aliases=['greek'], the 1st character p would be returned because this latin letter can be confused with greek ρ.

  • with preferred_aliases=[] and greedy=True, you’ll discover the 29 characters that can be confused with p, the 23 characters that look like a, and the one that looks like ρ (which is, of course, p aka LATIN SMALL LETTER P).

confusables.is_dangerous

confusables.is_dangerous(unicode_string, preferred_aliases=[])

Boolean: True if is_mixed_script(unicode_string) and is_confusable(unicode_string).

The preferred_aliases argument is simply passed to is_confusable.

Is the data up to date?

Yep.

The unicode blocks aliases and names for each character are extracted from this file provided by the unicode consortium.

The matrix of which character can be confused with which other characters is built using this file provided by the unicode consortium.

This data is stored in two JSON files: categories.json and confusables.json. If you delete them, they will both be recreated by downloading and parsing the two abovementioned files and stored as JSON files again.