Abaca ReceiverNet: Interesting Spam Control Technique

We were recently briefed by Abaca, a vendor of spam control technology, on its proprietary ReceiverNet spam control algorithm. Abaca makes impressive claims for the accuracy and performance of ReceiverNet. The underlying algorithms have been explained to us under NDA, and they are very interesting. In essence -- and at the risk of over-simplification -- it relies on understanding the relationships between senders and recipients.

The ReceiverNet algorithm seems to be a good way of automatically generating on-the-fly, per-user whitelists and blacklists, with minimal time delay. The technology employs a more rigorous, statistical approach to this, which we found impressive. It helps to differentiate between purely-spammy senders, and those that appear to send both spam and ham email (for example, where a good sender shares an IP address with a bad sender). This should help prevent false positives.

ReceiverNet also keeps track of what proportions of good and bad email a recipient usually gets. This effectively turns every user into a fuzzy spamtrap. This aggregated data is used to weight the spam/ham decision.

Of course, this idea won't work so well in the case when little or no data is known about the sender -- the "zero-hour" problem. However, ReceiverNet also does statistical analysis of the content of known spam and ham messages. This allows it to compare known content with the content of unknown messages.

If it's still not clear whether the message is spam or ham, ReceiverNet adds it to a separate inbox of "uncertain" messages. It's good to see that Abaca doesn't simply dump the such messages into a huge quarantine -- it's doing what I have often suggested spam control should do: prioritize the quarantine, to show the most likely false positives first.

Of course, no shiny, new spam control technique is perfect. Over the years we've seen many new techniques promise much and deliver little. But we like Abaca's approach, it's very promising.

... Richi Jennings, with Nick Shelness

Post a comment

You must be logged in to post a comment. To comment, first join our community.