[Labs-l] Efficient way to deal with wiki spam on Labs

Thu Jul 4 09:15:07 UTC 2013

Le 04/07/13 06:30, Matthew Flaschen a écrit :
> 3. Some kind of AbuseFilter/TitleBlacklist-based solution.  This seems
> promising, and I'm wondering if some other Labs wiki has experience with it.

Beta cluster has the same issue.  We got AbuseFilter rules maintained
centrally by Steinsplitter and others.  That seems to be quite
effective, I am wondering whether we could have a labs centralized
database to maintain those rules for all labs wiki.

The DNS blacklist does catch some open proxies referenced on sorbs.net,
it is configured on beta via wmf-config/CommonSettings-labs.php  :

$wgEnableDnsBlacklist = true;
$wgDnsBlacklistUrls   = array(
    'proxies.dnsbl.sorbs.net.',
);

That is really only a small part of possible spam attempts though.

If we had a centralized AbuseFilter instance for labs, we could publish
the blacklisted IP over DNS and have all labs wikis to use that DNS
blacklist.
We could even make it public for other people to use  :)

I used to blacklist Amazon instances as well.

There is a lot of logs on beta under /data/projects/logs, that might a
good way to find patterns / offending hosts.

-- 
Antoine "hashar" Musso