Yesterday Barracuda Networks announced an upgrade to its Spam Firewall product that provides the ability to apply optical character recognition, or OCR, techniques to incoming e-mail. This is an important escalation in the spam/anti-spam arms race, and something that will hopefully become more common.According to the press release, "as much as 25 percent of all spam messages contain images," and from a quick survey of my mail logs, that figure pretty well jibed with what I saw there. Heck, 7% of the mail to hit my server today contained an image and nothing else. I can punish the GIF-only mail easily enough with regular expressions, but the mixed-mode spam is much more troublesome. Fortunately, techniques like the one announced by Barracuda should help alleviate some of these problems.
It'll be interesting to see how well this works. This is something that has been talked about in various anti-spam groups for a while, but as far as I know nobody has productized it until now. I'm going to assume that Barracuda has done some internal testing on this already and has determined that the processing demands aren't terribly excessive--and since Barracuda relies largely on SpamAssassin and other open-source technologies, the cost burden is probably next to nil, too--so we ought to start seeing this appearing elsewhere in short order.