@Makyen Geez. Out of curiosity, do you have a tool that generates these? Or do you just use a list of all the different characters, from which you copy-paste to create the regex manually?
@CodyGray I generated that manually from the list of character equivalences provided on Homoglyph Attack Generator. The lists there are substantially incomplete, but are a reasonable starting place for something quick (e.g. for "0" it has 13 equivalents, but the number_homoglyphs.py module uses 133 equivalents for "0", which still isn't complete). As I recall, a better starting point is here.
After having generated that watch entry, I've been thinking (again) about various ideas for automatically handling such things, similar to how the number_homoglyphs.py module handles homoglyphs/deobfuscation for numbers, and how to apply such at least semi-automatically to the various lists, along with other options/improvements. There are a lot of interrelated things to do in that area of the code.
In other words, creating that was manual, with substantial possibilities for automation.
In general, we should probably avoid having individual regular expressions which attempt to each perform deobfuscation to that extent. Overall, it's inefficient to do it that way and difficult to maintain. I chose to do it that way in this case, due to the increasing number of posts for that particular spam and their demonstrated willingness to obfuscate in that manner.
In general, spammers aren't willing to obfuscate things like URLs, because doing so effectively requires that people that see the spam have to put out substantially more effort to actually go to the URL and the spammer doesn't get anywhere close to the SEO benefit as they would if they didn't obfuscate to that extent. Thus, handling such deobfuscation in a general programmatic manner has been a lower priority, just because it's not used by spammers nearly as often as is used for phone numbers.
@CodyGray That pattern looks like it's already caught by Potentially bad keyword in answer and Potentially bad keyword in body Append -force to the command word(s) if you really want to add the pattern you provided.
@SmokeDetector tpu- ChatGPT-generated, with possible spam link (unclear whether that's actual spam, or someone trying to be "helpful" by using ChatGPT); custom-flagged for mod attention.
@CodyGray That pattern looks like it's already caught by Potentially bad keyword in answer and Potentially bad keyword in body Append -force to the command word(s) if you really want to add the pattern you provided.
Any ideas about this suggested edit? askubuntu.com/review/suggested-edits/1273153 The description contains that number, but that format appears to be for TV/movies (I.e., not real). The edit removes formatting but seems overall odd
@SmokeDetector tpu- Contains spam link, text excluding link is 99.10% ChatGPT per the huggingface detector. Custom flagged
Oh, I see. Never mind. Yeah, I've no idea about that suggested edit. Sometimes you see weird stuff suggested.
Why does it not say what user suggested it? Shouldn't it say that it came from an anonymous user if it did, rather than just omitting the usercard altogether?
!!/bisect work up to bigger stuff like car batteries. He keeps the source of the batteries to himself, telling his distributors he's smuggling the batteries
@SurajRao Not a link, just "Introduce the problem and expand on what you put in the title. Minimum 20 characters." Copy-pasta gibberish.
@CodyGray The pattern contains an invisible U+202D whitespace character. Append -force to the command word(s) if you really want to add the pattern you provided.
@CodyGray That pattern looks like it's already caught by Potentially bad keyword in answer and Potentially bad keyword in body Append -force to the command word(s) if you really want to add the pattern you provided.
I would just call it "youtube" in the inline regex comment. As I understand it, the URL shortener youtu.be uses the same unique IDs as the main site.
Obviously, this is not a big deal, but it might make it easier to search for all items on the list that were YouTube video IDs (if we ever wanted to do that).
@cocomac FYI: Typically, in the North American Numbering Plan (NANP), the 555 prefix (central office code, not area code) is almost always reserved with the only functional number being (area code)-555-1212, which is directory assistance. See 555_(telephone_number) for more information. It's quite unlikely that a spammer would use a phone number with a 555 prefix, but a troll might or someone testing.
