Defensio - Outsmartting Evil SpamIn my post about preventing comment spam on blogs last month, I mentioned that I signed up to be a beta tester for a new anti-spam filter that promises to be better than Akismet. This is not to say that Akismet is bad, on the contrary it is very good but it does make one or two false positives a week (real comments that get flagged as spam).

Defensio is a new spam filter developed by three guys from Montreal, Canada. I was one of the few blogs selected for the initial beta testing so I figured I should give my thoughts on it.

A couple days ago, I received an e-mail from the Defensio team with a link to the beta version of Defensio WordPress Plug-in, my unique API key and instructions on how to install it.

Installation
Installation is easy enough — it installs just like any other WordPress plug-in.

  1. Upload to plugin directory.
  2. Deactivate existing spam protection (Akismet, Bad Behavior etc.).
  3. Activate plugin.
  4. Enter API key.

How it Works
Using their own algorithm, Defensio assigns a Spaminess value to each comment ranging from 0-100%. Comments are then filtered based on their spaminess value and those that are marked as spam go into the Quarantine. I’m happy to report that this spaminess statistic appears to be working. Although I’ve gotten some false positives, they were the lowest in spaminess value in the quarantine.

On a side note, I took a look at the source code and it’s very well organized and documented — kudos to the Defensio team for that!

User Interface
The interface is very nice and much better than Akismet’s (but similar to Akismet with Auntie Akismet script). Comments are trimmed to a single line to save space but you can expand the comment to see the full text. Restoring comments uses a little AJAX goodness complete with prototype goodness. I know some people don’t see much value in this but I think it makes the otherwise boring task of comment moderation just a little bit more enticing.

Defensio Quarantine

Also, the quarantine auto-hides obvious spam which are comments with a spaminess value greater than a customizable threshold. You can choose to display this obvious spam but so far, everything truly has been obvious spam. As an added touch, comments are shaded with a color corresponding to their spaminess value — the redder the badder.

Does it work?
So far it has missed a couple spam comments and mistakingly marked a few real comments as spam but I feel it’s too early to get a good picture of its true potential. Unfortunately, I can’t run it in conjuction with Akismet to see which one filters more accurately. I’m sure the team is currently tweaking and improving it so I’ll let it run for a few weeks to see how things work out.


Popularity: 2% [?]

Leave a Reply