Wikipedia:WikiProject Spam/IRC Channels

Some members of the Wikiproject on Spam are active on IRC, where part of the control and detection of spam is located. In these channels several programs are active which monitor and revert external links additions

Connect to IRC
To connect to the channels you need an IRC client. See IRC/Tutorial for a list of popular clients.

Channels

 * #wikipedia-spam-t is the main talk channel. Some of the bot-accounts can be controlled from here, and they produce some warnings here, but this channel is mainly used for discussions.
 * IRC channel:


 * #wikipedia-en-spam is the main channel for the bots for the English wikipedia
 * IRC channel:


 * #cvn-sw-spam is the channel for the bots where there is no specific channel.
 * IRC channel:


 * The main effort of cross-wiki vandalism is in #cvn-sw, some of the bots also generate warnings for the non-english wikipedia here.
 * IRC channel:

General
Most of the bots work on the basis of regex (regular expressions). This format is also used on the blacklists and on XLinkBots revertlist. When in doubt, please ask! Adding a 'bad' regex can have serious effects on the workings of the bots.

Bots
The following programs are active in the channels:

The LinkWatchers
These are our main bots. For every edit performed they determine if it is in one of the namespaces that it monitors, and if so, it fetches the diff from the server, finds in the diffs the links that were removed and added, and then reports which links were added (but not removed; hence all new links).

Bots

 * English language: 'EnLinkWatcher2' (English language wikipedia) on #wikipedia-en-spam
 * All other languages: 'BigWikiLW2' (the big ones, except English language wikipedia), 'WiktLW2' (all wiktionaries) and 'IntLinkWatcher2' (about 650 remaining wikipedia) on #cvn-sw-spam

These bots are run by en:User:Beetstra on a box owned by en:User:Versageek (both can start/restart the bots when necessary). Bots have been developed by en:User:Shadow1 and en:User:Beetstra. Since Shadow1 has retired, the bots are maintained by Beetstra only.

The bots read the diffs from in total 722 wikis on irc.wikimedia.org (list of a couple of months old). The diffs are parsed and external links that are added are reported by the bots.

If you need a bot for a specific language in a specific channel, please contact en:User:Beetstra. I am currently unable to fork of any language specific bots due to limitations in IRC/computer.

Commands
The bots share the database, commanding one will result in being added to the others as well.


 * Whitelist is for 'good' links
 * RedList for links which need attention
 * Blacklist for bad links (on #wikipedia-en-spam these are transferred to XLinkBot and reverted, except when the user is whitelisted). This blacklist is stored on-wiki and can be found here: User:XLinkBot/RevertList

Standard the settings are to hide whitelisted links and whitelisted users.

General
The data is stored in a database. If you need specific info, ask en:User:Versageek, en:User:Beetstra or en:User:Eagle 101, all three have access to the database.

XLinkBot
Reads reports from the linkwatchers and reverts blacklisted links by IP-users and unestablished accounts.

BetacommandBot3
Basic linksearch and crosswiki search tool. To give the bot a command use: BetacommandBot3  

Commands
 * parameters
 * Linksearch
 * website
 * betacommandbot linksearch blog.myspace.com
 * Linksearch2
 * website
 * betacommandbot linksearch2 blog.myspace.com
 * Crosswiki
 * website
 * betacommandbot crosswiki blog.myspace.com
 * crosswiki2
 * website
 * betacommandbot crosswiki blog.myspace.com
 * WPSPAM
 * betacommandbot WPSPAM
 * Linkfilter
 * betacommandbot linkfilter
 * linkgenupdate
 * betacommandbot linkgenupdate
 * blsite
 * rmbluser
 * rmblsite
 * showuser
 * showsite
 * blacklistupdate
 * wikilinksearch
 * add_access
 * del_access
 * save

COIBot
The IRC part of this bot is split into three bots, 'COIBot', 'LWCOIBot', and 'COIToolBot'.

How to react to IRC warnings
The bots generate a variety of warnings. Though it is impossible to react to everything, they are good starting points for some further investigation. The linkwatchers are the best starting point, as the other bots derive their data from them.

In all cases BetaCommandBot3 and the COIBots can be used for specific reports which can help to investigate links and/or users better (for commands vide supra).

LinkWatchers
The linkwatchers generate output for each edit where link additions are performed. Within its output there are some points which should alert:

Output
The bot gives one line per edit in the format '(pagename) (diff) (username) (list of links)', e.g.:


 * Example en:User:Example http://www.example.com/ ( 4, 3, 2, 1) (BL: \bexample\.com) http://www.example.com/ ( 4, 3, 2, 1) (BL: \bexample\.com)

Every link in the list is followed by a set of 4 numbers ('( 4, 3, 2, 1)'), and may be followed with a warning if the link is on the white, red or blacklist. The four numbers:


 * 1) how many links did this user add.  Turns red when it is higher than 3.
 * 2) how many times was this link added.  Turns red when it is higher than 250 (can be considered for whitelisting)
 * 3) how many times did this user add this link.  Turns blue when that is higher than 5, turns red when higher than 25.  The number may be followed by "!!" (in red), in that case more than 90% of this link was added by this user, "!" in red (more than 66%) or by "?" (more than 33%).
 * 4) to how many wikipedia did this user add this link.  Turns blue when higher than 2, turns red when higher than 5.  The number may be followed by "!!" (in red), in that case more than 90% of this link was added interwiki by this user, "!" in red (more than 66%) or by "?" (more than 33%).

Todo:
 * Check users which have added a high number of links (first number):
 * If it is a regular editor, whitelist the editor ('EnLinkWatcher2 user add ')
 * If a spammer, check edits and react accordingly.
 * Check links which have been added a lot (second number):
 * Good link, consider whitelisting ('EnLinkWatcher2 link wl add ').
 * Otherwise, leave as is, or consider reporting to e.g. WT:WPSPAM.
 * When one of the last two numbers is high in comparison to the first two, the user is adding the same link quite often. It may still be a good link, but chances are that the user is spamming.

Automatic alerts
When a user has a focus for a certain link (either being the only user adding the link, or adding a link cross-wiki, then it will be automatically added to the redlist, and a warning will be given by the bot:


 * WARNING: Link \bexample\.com added to redlist, en:User:Beetstra is only user adding this link.

Todo:


 * Good link, consider whitelisting ('EnLinkWatcher2 link wl add ').
 * Otherwise, leave as is, or consider reporting to e.g. WT:WPSPAM.

Redlist
The linkwatchers redlist links that have been added cross-wiki by one user, or when one certain user has a large preference for using this link. When you have concerns about a link, but don't have the time to investigate, then you can also add a regex to the redlist yourself: 'EnLinkWatcher2 link rl add '

Todo
 * When a linkwatcher reports a redlisted link:
 * Have a look at the edits of the editor adding this occasion.
 * Have a look at the site being added.

Then:


 * If the user is clearly spamming, but the site is OK, warn the user and try to engage in discussion
 * If the user is clearly spamming for commercial reasons, then revert the users, and warn the user for spamming
 * Consider either using User:XLinkBot to autorevert when the user persists.
 * When blacklisting has occured (the linkwatcher will show it), you can remove the link from the redlist
 * If the site is really good, add the link to the whitelist on a linkwatcher (e.g. 'EnLinkWatcher2 link wl add ' (preferably making a regex of the link) and remove the link from the redlist ('EnLinkWatcher2 link rl del ').
 * If the user is OK (as in, not a spammer or similar), consider whitelisting the user ('EnLinkWatcher2 user add '; username has to be exact with capitals and spaces)
 * wikipedia BOT-accounts: when a regular bot, preferably confirmed by WP:BAG, whitelist ('EnLinkWatcher2 user add '; username has to be exact with capitals and spaces)

XLinkBot
There are typically four warnings from User:XLinkBot:


 * "I have already reverted"
 * Check the edit and revert again when necessary, or leave when this was a good edit.
 * It is very important to check the edit here, please do not automatically revert just because someone reverted the bot.


 * Previous revert was by another AntiVandalismBot
 * Check the edit. XLinkBot does not revert to Cluebot and VoABot II
 * It is very important to check the edit here, please do not automatically revert just because someone reverted the bot.


 * User added link inside a reference, template or remark
 * Check the edit.
 * User does not satisfy autoconfirm
 * Check the edit. This is a user who has been registered on Wikipedia for more than 7 days. While longevity does not guarantee a user is familiar with our policies, the longer a user has been editing, the more likely they are to understand policy.

XLinkBot is giving some other errors sometimes, just check the edit that XLinkBot is referring to.

General
The last part in the posts by COIBot and LWCOIBot ('Report 1/2') gives the occurance of the edits by this editor, the first number is how many edits this editor has done to pages where they may have a conflict of interest in the last hour, the second number his/her total number of edits where they may have a conflict of interest. When these numbers are relatively high, then these edits need to be investigated.

COIBot
This bot only reports overlaps between username and pagename:


 *  (Overlap) calculated overlap Beetstra <-> Beetstra ( http://en.wikipedia.org/w/index.php?title=Cheap_Trick&diff=202544343&oldid=202543883 ; 100.0%) (report 1/2)
 * Todo:
 * Check the page:
 * When clearly commercial or promotional, delete or nominate for deletion.
 * When not commercial, but editor clearly has a conflict of interest and that is noticable from the page, then consider tagging the article with coi or coi2, or reporting the editor to the conflict of interest noticeboard
 * Wrong overlap, whitelist the user against the page ('COIBot wl add '; replace spaces in username and pagename with an underscore (so 'Conflict of interest' becomes 'Conflict_of_interest'), or for regular editors, consider whitelisting the user against everything ('COIBot wl add *'; again replace spaces in username with an underscore).

LWCOIBot

 *  (Overlap) calculated overlap Beetstra <-> Beetstra.com (100.0%) (report 1/2)
 * Someone added a link which is very similar to the username
 * Todo:
 * Check the edit and the link added.
 * When the added link is spam, revert and warn. Consider blacklisting or adding to XLinkBot's revertlist when user is persistent.
 * When link is OK, consider engaging in discussion with user. Often it is appropriate to contact a wikiproject.
 * Wrong overlap:
 * Whitelist the user against the url ('COIBot wl add '; replace spaces in username with an underscore (so 'Conflict of interest' becomes 'Conflict_of_interest'), or for regular editors, consider whitelisting the user against everything ('COIBot wl add *'; again replace spaces in username with an underscore).
 * The link has probably also been put on the monitorlist, please remove the link from that list ('LWCOIBot ml del ').


 *  (Monitor) eu:user:Beetstra added www.example.com in (rule: \bexample\.com - reason: Automonitor: COI? (calculated overlap Example <-> Example.com; ); report 1/1
 * Someone added a link which matches a rule on the monitorlist
 * Todo:
 * Check the edit and the link added.
 * When the added link is spam, revert and warn. Consider blacklisting or adding to XLinkBot's revertlist when user is persistent.
 * When link is OK, consider engaging in discussion with user. Often it is appropriate to contact a wikiproject.
 * Wrong overlap:
 * Whitelist the user against the url ('COIBot wl add '; replace spaces in username with an underscore (so 'Conflict of interest' becomes 'Conflict_of_interest'), or for regular editors, consider whitelisting the user against everything ('COIBot wl add *'; again replace spaces in username with an underscore).
 * The link has probably also been put on the monitorlist, please remove the link from that list ('LWCOIBot ml del ').

The remarks: are for information only and can generally be ignored. Note that when links get reported they are picked up by LWCOIBot and monitored.
 *  Stalked page en:MediaWiki talk:Spam-blacklist edited by en:User:Beetstra, checking for added LinkSummary templates.
 *  Stalked page en:Wikipedia:Conflict of interest/Noticeboard edited by en:User:Beetstra, checking for added LinkSummary templates.
 *  en:MediaWiki talk:Spam-blacklist edited, no LinkSummary templates added.
 *  Stalked page en:MediaWiki:Spam-blacklist edited by Beetstra, checking for added links.
 *  en:Wikipedia talk:WikiProject Spam edited, COIBot automonitored urls: \bexample\.com
 *  en:MediaWiki:Spam-blacklist edited, COIBot automonitored urls: \bexample\.com\b

COIToolBot
This bot can generally be ignored, it is just reporting when it saves reports. It does report to the talkchannel after saving a cross-wiki or requested report.