flock2016
LOGS
14:39:10 <nardasev> #startmeeting <How we took care of Spam>
14:39:10 <zodbot> Meeting started Wed Aug  3 14:39:10 2016 UTC.  The chair is nardasev. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:39:10 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
14:39:10 <zodbot> The meeting name has been set to '<how_we_took_care_of_spam>'
14:39:19 <nardasev> #meetingname flock2016
14:39:20 <zodbot> The meeting name has been set to 'flock2016'
14:39:43 <nardasev> the beginning: started with some wiki spam
14:39:51 <nardasev> the fight begins:
14:39:59 <nardasev> range and speed increased a lot
14:40:05 <nardasev> trac was added to the mix
14:40:17 <nardasev> getting out of manual control levels
14:40:27 <nardasev> temporary CLA+1 wiki and trac
14:40:33 <nardasev> first scripts went live
14:41:23 <nardasev> ./watch_edits.py | ./is_spam.py | ./deal_with_spam.py
14:42:08 <nardasev> only people who are contributing to fedora could log in to wiki
14:42:24 <nardasev> at one point it became ridiculous: hundreds of accounts a day
14:42:28 <nardasev> Basset begins
14:42:42 <nardasev> restructured the fight against spam from scratch
14:43:03 <nardasev> lots of plugins for MediaWiki to fight spam, but not as many for Trac
14:43:13 <nardasev> wish to consolidate all spam fighting
14:43:59 <nardasev> basset is now getting support for other things a sother sys admins are getting under attack
14:44:06 <nardasev> *as other
14:44:13 <nardasev> showing chart
14:44:32 <nardasev> Wiki, FAS, and Trac go to Basset frontend
14:44:45 <nardasev> basset frontend goes to Basset workers
14:45:06 <nardasev> Basset workers go back to Trac and Wiki
14:45:38 <nardasev> Basset gets messages from Wiki, Trac, FAS, Pagure, and determines a score based on certain modules, like spam words
14:46:39 <nardasev> based on the final score it will either happily accept your message, or delete your message and account, or if it's not sure, it will send a message to the admin
14:46:46 <nardasev> we are getting better
14:49:03 <nardasev> as soon as you start editing a wiki page, it will send a message to Basset with the information what you entered and usernam
14:49:13 <nardasev> it will again give you a score
14:49:46 <nardasev> you'll either get blocked on FAS account or accepted
14:50:02 <nardasev> it mostly works, but sometimes things slip through
14:50:11 <nardasev> so we edit manually
14:50:28 <nardasev> we've had both false positives and false negatives
14:50:55 <nardasev> lots of training and teaching has to happen
14:51:06 <nardasev> more ML plugins coming into production
14:51:17 <nardasev> more projects looking at deploying
14:51:30 <nardasev> increasing contributor and user base
14:52:45 <nardasev> I'm very passionate about reducing spam
14:53:11 <nardasev> wiki became so ridiculous with spam that at one point we decided to make wiki CLA+1 for the time being
14:53:24 <nardasev> I hope we can get rid of the limitation
14:53:39 <nardasev> 3000 new accounts a day, because an account gets blocked as soon as it
14:53:46 <nardasev> is classified as spam
14:53:50 <nardasev> Future?
14:54:17 <nardasev> Couple of plans in motion to also resolve this issue, which will be launched over tim
14:54:20 <nardasev> *time
14:54:32 <nardasev> if you have any suggestions, I will accept anyone's help
15:01:44 <nardasev> Q: does basset also work as an anti spam service for email server?
15:02:05 <nardasev> A: we are aiming for that once we have synchronized API
15:04:42 <nardasev> #endmeeting