Page MenuHomePhabricator

Spam filter for the Konkani WIkipedia
Closed, InvalidPublic

Description

In the past few months, there are spam posts being created in Russian regularly on talk pages on the Konkani Wikipedia (on the Konkani Wikipedia, the talk pages are in Structured Discussions - formerly Flow - format by default). I am a temporary admin on gomwiki and I delete these posts as soon as I see them. I would like to request help to create a local abuse filter for this.

I have created a page with links to instances of spam that I have deleted, and you can also take a look at the deletion log for the last few months.

I have studied the instances of spam, and I think that the filter should have the following conditions:

  • Filter group: Flow
  • user_age==0 (Unregistered editors)
  • board_namespace==3 (User talk pages)
  • It should check that at least 40% of the text should be Cyrillic characters

It appears that we are dealing with real people and not bots, therefore, warnings will not be sufficient, and the edit should be blocked.

I have also noticed that at least a couple of the IP addresses have been reported as forum spammers, so perhaps later on, if necessary, we can create a separate filter using sfs_blocked that checks whether the IP address is blocked using the stopforumspam.com list and more stringent criteria regarding the contents.

Since I'm not well-versed with writing abuse filters, please could someone help by writing a local filter with the above conditions? If you could get us started, I'll try to maintain the filter myself later on.

Event Timeline

@The_Discoverer: As this does not seem to be a topic about actual AbuseFilter code, but about its configuration on a wiki, https://www.mediawiki.org/wiki/Extension_talk:AbuseFilter or https://www.mediawiki.org/wiki/Extension_talk:AbuseFilter/Rules_format might be more appropriate venues?

Also CC'ing @Daimona for a potential opinion ^

(Personal question out of curiosity: What would be the one documentation page you'd link people to, to learn the basics of writing AbuseFilter rules?)

Daimona claimed this task.

@The_Discoverer: As this does not seem to be a topic about actual AbuseFilter code, but about its configuration on a wiki, https://www.mediawiki.org/wiki/Extension_talk:AbuseFilter or https://www.mediawiki.org/wiki/Extension_talk:AbuseFilter/Rules_format might be more appropriate venues?

Yup, https://www.mediawiki.org/wiki/Extension_talk:AbuseFilter is usually the right place for these questions.

The rules you wrote in the task description seem reasonable, too. Maybe it could be a little tricky to get the last one right (40% of the text is cyrillic), but I think it should be doable with rcount + strlen. Also, note that StopForumSpam is not enabled on WMF wikis, so you won't be able to use sfs_blocked.

To write a filter, you have to go to Special:AbuseFilter, hit "Create a new filter", and put the conditions there.

(Personal question out of curiosity: What would be the one documentation page you'd link people to, to learn the basics of writing AbuseFilter rules?)

Hmmm, good question.

  • This paragraph on mw.o gives a very basic overview of the available help pages;
  • /Rules format documents the whole filter language, variables, etc.
  • /Actions describes the actions available.
Aklapper changed the task status from Resolved to Invalid.Feb 16 2020, 4:43 PM

Thanks for the quick answer! Marking this task as invalid as no code was changed in the AbuseFilter code base.