Right to left override messes up the text in reviews

Bug
1

The image shows a content moderation report detailing flagged words in a post, including various exclamations, slurpeh, and other partially visible words, which require approval. (Captioned by AI)
The image shows a content moderation report detailing flagged words in a post, including various exclamations, slurpeh, and other partially visible words, which require approval. (Captioned by AI)1159×374 102 KB

When a left to right override is added , it will mess up the review queue text. Version of forum is 3.5.0.beta2-dev

1 Like
2

Can I clarify what you mean here? Are you referring to a specific site setting?

3

I’m talking about the emoji itself, not anything else: U+202E RIGHT-TO-LEFT OVERRIDE - Unicode Explorer
If the Unicode character is put into the flag reason, it will break the text.

Like this (reversed):
‮like this

5

Thanks for reporting this @twofoursixeight. Judging by the effect that this unicode character has on text, the flag reason is showing as expected there, even if it does appear garbled - it’s showing the text as it was typed by the user. It also appears that the watched words are being flagged correctly, despite the backwards writing, so I don’t know if this is a bug as such.

Aside from the text appearing backwards in the flag reason (which is as expected, given the existence of the unicode character), is there anything else problematic here?

6

The right to left override can also be used to bypass automod, see this:

no right to left override Unicode:

This image displays a text-based interface where "Fuck" is entered for checking against a list of "watched words," resulting in "Fuck" being found as a match. (Captioned by AI)
This image displays a text-based interface where "Fuck" is entered for checking against a list of "watched words," resulting in "Fuck" being found as a match. (Captioned by AI)1272×684 103 KB

with the right to left override (notice the grammar check? autocorrect is thinking of a different word!

The image shows a blocked word alert notification with the word "Fuck" currently being monitored and a message stating no matches have been found. (Captioned by AI)
The image shows a blocked word alert notification with the word "Fuck" currently being monitored and a message stating no matches have been found. (Captioned by AI)1258×569 103 KB

What is happening here: the watched word is shown as f*ck which will be flagged by the moderators. However, if you put a Right to Left override before it, you need to type in reverse, meaning you have to type this instead: kc*f
The watched words filter will not recognize words correctly thanks to the override.

1 Like
7

Got it - that makes sense, and I can see how that would be a way for people to work around spam filters.

As a workaround, while we figure out the nest way to handle this in core, if you add that unicode character as a watched word, it will get flagged in posts. Just go the page you linked, hit the copy button there, and paste that in as a new watched word to flag. That will cause any post using it to be flagged for review. I just tested that and it works correctly.

While there are legitimate use cases for using it (like adding some text from an RTL language inside other text), it is certainly a potential spam vector. Adding it as a watched word will help you identify when it is used and review the posts as they come in. In the meantime, we’ll see what would be the best way to handle this in core.

1 Like