I saw a talk at hushcon about the problem of IDN domains and punycode
domains and the difficulty of covering every possible base.. I dont
really remember, but I liked the idea of trying to solve this analog
problem a bit with code and heres the PoC I wrote.
The idea is pretty simple, the problem with IDN characters is that
some characters may look similar to other caracters so OpenCV seemed the
natural solution to brute force possible lookalike options using different
alphabets and fonts, creates a score for how well each character matches
the targets and displays them cutoff by the threshold.
$ python eyephish.py --inputstring "amazon" --dialect=greek
*amazon ['a', 'm', 'a', 'z', 'o', 'n']οη [' ', ' ', ' ', ' ', u'\u03bf', u'\u03b7']
όή [' ', ' ', ' ', ' ', u'\u03cc', u'\u03ae']
ϙ [' ', ' ', ' ', ' ', u'\u03d9', ' ']
σ [' ', ' ', ' ', ' ', u'\u03c3', ' ']
ρ [' ', ' ', ' ', ' ', u'\u03c1', ' ']
ͻ [' ', ' ', ' ', ' ', u'\u037b', ' ']
ͽ [' ', ' ', ' ', ' ', u'\u037d', ' ']
ϱ [' ', ' ', ' ', ' ', u'\u03f1', ' ']
ϼ [' ', ' ', ' ', ' ', u'\u03fc', ' ']
>$ python eyephish.py --inputstring "microsoft" --dialect=cyrillic
>*microsoft ['m', 'i', 'c', 'r', 'o', 's', 'o', 'f', 't']
>> ісѓоѕоӏӏ [' ', u'\u0456', u'\u0441', u'\u0453', u'\u043e', u'\u0455', u'\u043e', u'\u04cf', u'\u04cf']
>> ӏҫґӧзӧӀӀ [' ', u'\u04cf', u'\u04ab', u'\u0491', u'\u04e7', u'\u0437', u'\u04e7', u'\u04c0', u'\u04c0']
>> Ӏҁ ө өІІ [' ', u'\u04c0', u'\u0481', ' ', u'\u04e9', ' ', u'\u04e9', u'\u0406', u'\u0406']
>> Іє ӫ ӫгг [' ', u'\u0406', u'\u0454', ' ', u'\u04eb', ' ', u'\u04eb', u'\u0433', u'\u0433']
>> ї е етт [' ', u'\u0457', ' ', ' ', u'\u0435', ' ', u'\u0435', u'\u0442', u'\u0442']
>> Ї р ріЇ [' ', u'\u0407', ' ', ' ', u'\u0440', ' ', u'\u0440', u'\u0456', u'\u0407']
>> г б бЇі [' ', u'\u0433', ' ', ' ', u'\u0431', ' ', u'\u0431', u'\u0407', u'\u0456']
>> т ё ёїӷ [' ', u'\u0442', ' ', ' ', u'\u0451', ' ', u'\u0451', u'\u0457', u'\u04f7']
>> ғ ѐ ѐТҭ [' ', u'\u0493', ' ', ' ', u'\u0450', ' ', u'\u0450', u'\u0422', u'\u04ad']
>> ӷ ҏ ҏ ї [' ', u'\u04f7', ' ', ' ', u'\u048f', ' ', u'\u048f', ' ', u'\u0457']
>> Т [' ', u'\u0422', ' ', ' ', ' ', ' ', ' ', ' ', ' ']
>> ҭ [' ', u'\u04ad', ' ', ' ', ' ', ' ', ' ', ' ', ' ']
>$ python eyephish.py --inputstring "ebay" --dialect=cyrillic
>*ebay ['e', 'b', 'a', 'y']
>>ерау [u'\u0435', u'\u0440', u'\u0430', u'\u0443']
>>ёо ӱ [u'\u0451', u'\u043e', ' ', u'\u04f1']
>>ѐҏ ӳ [u'\u0450', u'\u048f', ' ', u'\u04f3']
>>өӧ ӯ [u'\u04e9', u'\u04e7', ' ', u'\u04ef']
>>ӫ ў [u'\u04eb', ' ', ' ', u'\u045e']
>>ӗ [u'\u04d7', ' ', ' ', ' ']
>>о [u'\u043e', ' ', ' ', ' ']
>>ӧ [u'\u04e7', ' ', ' ', ' ']
Its also worth mentioning that it can do this trick without resorting to a foreign alphabet as well
$ python eyephish.py --inputstring "microsoft" --dialect=latin*microsoft ['m', 'i', 'c', 'r', 'o', 's', 'o', 'f', 't'] >microsoft [u'm', u'i', u'c', u'r', u'o', u's', u'o', u'f', u't'] > l e etf [' ', u'l', ' ', ' ', u'e', ' ', u'e', u't', u'f'] > I b bll [' ', u'I', ' ', ' ', u'b', ' ', u'b', u'l', u'l'] > ! p pII [' ', u'!', ' ', ' ', u'p', ' ', u'p', u'I', u'I'] > | !! [' ', u'|', ' ', ' ', ' ', ' ', ' ', u'!', u'!'] > : ii [' ', u':', ' ', ' ', ' ', ' ', ' ', u'i', u'i'] > f || [' ', u'f', ' ', ' ', ' ', ' ', ' ', u'|', u'|'] > t [ [' ', u't', ' ', ' ', ' ', ' ', ' ', u'[', ' '] > + T [' ', u'+', ' ', ' ', ' ', ' ', ' ', u'T', ' '] > [ [' ', u'[', ' ', ' ', ' ', ' ', ' ', ' ', ' '] > . [' ', u'.', ' ', ' ', ' ', ' ', ' ', ' ', ' '] > ; [' ', u';', ' ', ' ', ' ', ' ', ' ', ' ', ' ']
This script depends on OpenCV and PIL modules
todo: adding two character brute forcing since vv->w.. which opened up more complicated problems then im interested in solvoing for now.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent