It's not easy for the spam filters to decide if an e-mail is spam or not. We, on the other hand, have no difficulties recognising unwelcome commercial messages, but it's much harder for the machines.
That's why we have focused on testing how good four good hardware solutions are at keeping the spam out of the users' inboxes, of course without stopping any real e-mails from getting through. The latter is known in the business as false positives. From the users standpoint it's just as important that real e-mails aren't caught in the spam filter as it is to get less actual spam.
12,000 spam mails in five weeks
To control how well the manufacturers' own calculations agree with reality we collected spam during a period of five weeks in four normal Swedish e-mail inboxes. Every kind of spam filter was turned off and we sorted the mail manually.
We separated the spam into categories: 1) written in English, 2) using Cyrillic lettering, 3) written in Swedish. During the five weeks we prepared for the test we managed to catch 11,380 spam messages, and we have used those on the tested products.
To check if they managed to let real e-mails through we used 2,140 different mails, both regular mail with conversations in Swedish and English, and a number of subscriptions to lists as well as mails advertising something. None of these mails were constructed in attempts to "nail" the filters; we have only used completely authentic material.
Two different technologies
Participating in the test is Symantec Brightmail 8260 and Ironport C300, both are using the same Symantec technology, which goes by the name Brightmail. The thought behind Brightmail is that it shouldn't need to resort to guessing games when incoming mail is classified as spam. Symantec has tens of thousands of mailboxes around the world, all used for collecting spam and legitimate e-mail. Every mail is sent to Symantec for control and verification: is it new spam in circulation or a legitimate mail?
Läs artikeln på svenska
