Last modified: 2014-04-27 17:18:41 UTC
The spambot's regexes currently are extremely out of date (1+ years old) and a majority of the warnings are false or reasons for the matching of 'spambots' is literally two closely named accounts. Task list; [ ] Check if the current regexes have any value anymore [ ] Update it with new spambots or LTAs. AbuseFilters may help here.
More ideas: * AbuseLog * Steal data from #cvn-sw-spam (LiWa and COIBot) * Users/IPs creating other users' talk pages * User-definable regexes, etc. (depends on bug 60858) * Same page name or edit summary being used xwiki
(In reply to comment #1) > * Users/IPs creating other users' talk pages Meant to say "Users/IPs creating other users' user pages"
(In reply to comment #1) > More ideas: > * AbuseLog > * Steal data from #cvn-sw-spam (LiWa and COIBot) > * Users/IPs creating other users' talk pages > * User-definable regexes, etc. (depends on bug 60858) > * Same page name or edit summary being used xwiki We should think about an automatic way to set these regexes. Or at least semi-automatic. It is indeed easy to know if a user is a spambot, as it is written in the lock reason entry. So the bot could learn himself if a user is a spambot or not, depending on used patterns and how he generally behaves.
(In reply to Quentinv57 from comment #3) > (In reply to comment #1) > > More ideas: > > * AbuseLog > > * Steal data from #cvn-sw-spam (LiWa and COIBot) > > * Users/IPs creating other users' talk pages > > * User-definable regexes, etc. (depends on bug 60858) > > * Same page name or edit summary being used xwiki > > We should think about an automatic way to set these regexes. Or at least > semi-automatic. > > It is indeed easy to know if a user is a spambot, as it is written in the > lock reason entry. So the bot could learn himself if a user is a spambot or > not, depending on used patterns and how he generally behaves. Adding 60858 as dependency since users should be able to add patterns manually as well. Good ideas.
