[haiku-web] spam checking routines

From: Matt Madia <mattmadia@xxxxxxxxx>
To: haiku-web@xxxxxxxxxxxxx
Date: Thu, 2 Sep 2010 21:11:18 +0000
This details some of my anti-spam & anti-spammer routines.

There's three types of spam that occur on the website:
 1. Backlink Accounts
 2. Comments & Replies
 3. Submitted Content

Each has a different routine for checking and dealing with.
In short, any spam postings & their respective spammer get deleted on the spot.

== Backlink Accounts ==

Backlink accounts are the most common type of spam I see.
It's from people who create a new user account for the sheer purpose
of having a place to spam a URL to their website (commercially
related). The URL's themselves are usually a dead giveaway -- eg,
www.buy-my-crap.com

A while back, Urias created a role "User Administration".
This role is essentially a full-fledge admin, as it also has the
permissions to modify the roles of other users. Additionally, this
role receives emails upon account creation & modification.

I've some email filters to help sort these emails, as it can be
voluminous. Later, I'll try to document them and store that
information on vmweb.

These accounts get deleted immediately, without hesitation.
If I notice that the same email address is being used, i'll sanitize
the account information & block the account (not the norm)
If I notice the same IP is being used to create accounts, I'll block
it. (not the norm)

== Comments ==

 * Published : http://www.haiku-os.org/admin/content/comment
 * Approval queue : -- not being utilized, the spammy comments will
simply be unpublished --
 * (Un-published) Spam : http://www.haiku-os.org/admin/content/comment/list/spam

Published comments from unfamiliar names get checked first, as it may
be spam that the filter didn't pick up.
For published spam, I don't bother "mark as spam" anymore -- i think
that caused the filters to become too aggressive as the spammers'
posts were too sneaky.
Un-published comments get checked to see if there's any non-spam that
was marked accidentally.

== Content ==

 * http://www.haiku-os.org/admin/content/node

In here, I'll filter  "status" is unpublished.
That will display forum posts (not replies)
and any other content that was submitted and detected as spam.


Another day, I'll try to remember to explain the different spam
filters used on the site.
... though, that may be better kept on vmweb too.

--mmadia
-----------------------------------------------------------------------
haiku-web@xxxxxxxxxxxxx - Haiku Web & Developer Support Discussion List
[haiku-web] spam checking routines

Other related posts:
