Posted on July 10, 2012
By Daniel Morrison
This morning we deployed changes to our spam checking code. We’ve been seeing steady increases in both spam comments, and false negatives. These false negatives meant many of us had lots of spam comments emailed to us and showing up on our blogs until we caught them.
On Sunday, we also saw a huge increase in comments coming in, and most of them were spam.
Today we’ve switched spam checking providers, and we’re comparing 3 side-by-side with code in-place to switch quickly if we find one performing better than the others.
We’ve also realized we didn’t have good admin-level insight into spam. We can see graphs like above, that show the amount of comments coming in, but not who is being affected and how our spam filters are performing. We’re working on improving those so we don’t have to rely on hearing feedback that spam is a problem. As an example, while one of our blogs was getting about 50 spam messages leading to a single url, a customer’s blog was getting 1,000+ with a hack trying unix commands. We plan to better see what’s going on in the system as a whole, and stop problems before they arrive in your inbox.
We feel the pain of spam as much as you do, and we’re happy to feel like we have it back under control. If you notice a difference, good or bad, please let us know.