Every day, literally hundreds of spam comments are sent to this blog. I have a a couple lines of defense, and generally they work pretty well. My first defense is a product called Bad Behavior, which inspects incoming messages and blocks the ones that look malicious before the WordPress code is even started up. Stopping evil at this stage can save a lot of server resources, as well as prevent this site from being hijacked by an unknown WordPress vulnerability.
Comments that get through that layer are then inspected to see if they look suspicious. Ones that the inspection service doesn’t like get thrown into a bucket behind the scenes where I can inspect them and approve innocent comments that were mistakenly flagged as spam.
I have been using Akismet for that, and in general I’ve been pleased with the results. The only downside is that now there are so many suspicious comments that I’m afraid that I’ll miss actual legit comments that were improperly flagged. Scanning through a list of hundreds of comments each day is not effective and, really, not a good use of my time. So, I began to look for alternatives.
Defensio is similar to Akismet, in that comments are shipped off to some service somewhere and then returned with a grade. The main difference is in the administration interface that I see, where Defensio sorts the rejected spam comments to allow me to more quickly spot legitimate comments that were falsely flagged as spam.
You may have noticed a surge in the amount of spam around here. This is (I hope) a learning phase for Defensio, and eventually it will stop allowing 3% of the spam comments to get through. (Akismet is still running, but mostly in a “see? I told you so” capacity right now.) I’m a little confused, because some of the comments Defensio displays are rated at 100% spamminess by Defensio’s own service.
Please bear with me through this somewhat-more-spammy-than-usual phase. I’ll be checking for spam comments regularly, and watching to see if Defensio’s performance improves. Also, this is a particularly good time to leave comments, from a training-the-filter perspective.