Extension talk:ConfirmEdit/FancyCaptcha experiments

Latest comment: 5 years ago by Nemo bis in topic 2019 discussion

userpages created by IPS

edit

From the mailing list, "2 userpages, which had not been previously created, were added by two IPs". Does that mean user pages were created by IPs? Seems like a reasonable scenario to throttle/block with AF..? That would be especially true on wikis with a Drafts extension, which is where IPs can?/should be able to.. post new pages. John Vandenberg (talk) 09:19, 5 December 2014 (UTC)Reply

Yes, I've now added that pattern to an existing filter. The only false positive in a month would have been a mistaken logged out edit; and it's an ongoing pattern with captchas enabled as well, so we should see some effect (and tweak accordingly) anyway. --Nemo 09:40, 5 December 2014 (UTC)Reply
Such a AF rule will be easy to create. I found the following:
and created this (currently disabled) AF rule for discusion: Special:AbuseFilter/38 (sorry, Nemo_bis was faster :/ --Florianschmidtwelzow (talk) 09:44, 5 December 2014 (UTC)Reply
Most users would see throttling/CAPTCHA instead of a logged out edit as a good thing! John Vandenberg (talk) 10:43, 5 December 2014 (UTC)Reply
Put me on the list of people in search of a big, bold, blinking "you're logged out again!" warning. OTRS would appreciate it. (Why can't MediaWiki detect that I was logged in a few minutes/hours/days ago, and prompt me to log in on the way to the edit page in the first place?) WhatamIdoing (talk) 01:09, 8 December 2014 (UTC)Reply

Recent experiments

edit

There might have been some new experiments after the one mentioned on this page, see phabricator:T158909#3225006. In general, the 2014 mediawiki.org experiment has shown that it's important to assess the results only in "realistic" conditions. In particular, certain repetitive patterns would naturally get blocked by AbuseFilters, which might not exist on the wiki being tested and should be included in global abusefilters. After the "obvious" global abuse filters are in place, any test can be repeated to give more realistic ideas of the amount of spam that FancyCaptcha specifically contributes to stop (which wouldn't be stopped otherwise). --Nemo 10:01, 3 May 2017 (UTC)Reply

2019 discussion

edit

I've mentioned this experiment in m:special:diff/19203243 (m:Wikimedia Foundation Board noticeboard). Nemo 18:57, 14 July 2019 (UTC)Reply

Return to "ConfirmEdit/FancyCaptcha experiments" page.