Last week, I installed the Defensio plugin on this blog to compare how Defensio handles comment spam with the standard functionality available in Habari.
Defensio is a centralised service aimed aimed detecting comment spam on blogs. To use Defensio, you firstly need to register in order to obtain an API key.
Installation of the Habari plugin is very straightforward and there are different versions available depending on your version of Habari.
- Subversion - http://trac.habariproject.org/habari-extras/browser/plugins/defensio/trunk
- Zipped archive for 0.7
- Tagged version for 0.6.2.
Documentation
The author of the Defensio plugin, Matt Read, documented the plugin on the Habari Wiki.
Configuration
Configuration of Defensio for Habari is very straightforward. You simply generate a Defensio API key and configure the plugin.
Usage
Although Defensio is a centralised service, only the validation is performed by Defensio and comments still appear on your blog and can be administered from the comfort of the Habari dashboard.
Defensio also allocates each comment a percentage ('spaminess factor') which is useful to assess how different comments are treated.
Defensio is more similar to Akismet rather than outsourced comment services (like Disqus, IntenseDebate) which also include some comment spam features and moderation is performed from the Disqus administration interface.
The Defensio plugin also includes a dashboard module which provides a quick summary of comments and spam levels.
Obviously, it's early days yet but so far, Defensio has correctly detected 8 comments as spam and 9 valid comments. It will be interesting to see whether Defensio does indeed train itself after some false negatives/positives have been flagged.
I've been using the Defensio plugin with both of my Habari installs for about 6 months - so far, there were only 3 false positives of about 600 spam comments (mostly in my "company" blog).
cu, w0lf.
Thanks - I was hoping that someone with longer, real-life experience of Defensio might pipe up.
I've been using it for probably a year now, and had 5318 spam comments, 347 innocent, 33 false negatives and 2 false positives. I'm having some trouble with my API key at the moment, but I also don't have much real traffic right now.