RegEx Is there a solution here?

Thomas 2020

I have a blacklist similar to this:

idiot(a|i|e)?
This block the idiot, idiota, idioti, idiote words.
But some users bypass this check by changing some letters with others
Ex: idi0t4
How can I make a regex list for block bypass attempts mimilar to this?

rinku singh

find what : idiot.

PeterJones

@Pan-Jan ,

Fighting blacklisted words is a difficult task. You have to consider all alternates, including letters, numerals, and other unicode symbols. There is no “magic” regex which will be able to recognize it for you.

For a generic hint, I suggest using character classes in square brackets [aie], not the groups-with-alternations (a|i|e). There is no reason to store the result in a group. It is easier to type the class without all the | separators.

Ignoring unicode (for now), a case-insensitive “idiot” regex might look something like (?i)[i|:!][d][i|:!][o0][t7][aie4]?, which says

case insensitive
things that look vaguely i-shaped
- | here is used as a literal character, because it looks like a capital i, not because it’s being used for alternation… another good reason to avoid the group where | has special meaning,
things that look vaguely d-shaped
things that look vaguely i-shaped
things that look vaguely o-shaped
0 or 1 things that look vaguely vowel-ending

Thomas 2020

But people have ideas.
That should probably be enough.

(?-i)[iіìílӏIƖІ][dԁDƊ][iіìílӏIƖІ][oοᴏOΟ0][tΤТƬ]

PeterJones

@Pan-Jan said in RegEx Is there a solution here?:

But people have ideas.

Indeed. I even had ideas, which I shared with you.

(?ii)

To me, it seems strange to turn off case-insensitive (with (?-i)), and then explicitly list both the lowercase and uppercase characters.

If I were trying to accomplish this, but for some reason couldn’t do the filtering through a command line script instead of doing it manually for each word inside Notepad++, I would at least use a command-line script to help write each the individual regex: I’d have the script ask for a word, like idiot, and then it would do an internal lookup from each letter to the list of characters that I thought were similar. (For example, that mapping would be i => "[iіìílӏIƖІ]", d => "[dԁDƊ]", ... according to your similarity rules.)

Good luck with this. Spam filters have been trying for years.

Thomas 2020

idiota
ɪdıota
iɗioti
ƖDI0T4
idiơte
iԁioti

There has to be a solution.

Can this be so?
[^ A-z1-9\n\r]|idiot

Thomas 2020

[^ A-z0-9\n\r]|idi[o0]t

Alan Kilborn

I like the test word chosen here.