The combined efforts of anti-spam products outperform any individual products alone, according to an experiment by Virus Bulletin, the independent security certification organisation.
In a comparative test, almost 200,000 sample emails were sent to 14 different anti-spam products that were required to filter out spam messages from legitimate smails (ham). The test found that no legitimate mail was blocked by more than four products.
The tests gave VB's anti-spam team the idea of a hypothetical spam filter that marked email as spam if at least five of the 14 products evaluated marked it as dodgy. Such a hypothetical filter would achieve a capture rate of 99.89 per cent and, better still, no false positives.
"For end-users this means that if spam filtering is business-critical, the use of more than one spam filter may be a good option," said VB's anti-spam test director, Martijn Grooten. "The anti-spam industry, meanwhile, should consider the benefits of collaboration and information sharing - and might be better able to protect our inboxes as a result."
The argument recalls the business case of firms such as GFI that have sold products that run two or more anti-virus engines in parallel to detect security threats against mail servers and the like. Running multiple products together makes up for the deficiencies of any single product providing, of course, the components lack similar shortcomings.
Security as a service firms, such as MessageLabs, run multiple spam and anti-virus filtering techniques to weed out junk or dangerous messages from email feeds. Grooten told El Reg he wasn't advocating any particular approach to spam filtering so much as encouraging more information sharing among anti-spam firms, which he reckons lags behind the sample sharing and collaboration commonplace among anti-malware researchers.
"Every spam filter uses multiple techniques but their effectiveness could be further improved if vendors share information on the latest spam attacks between each other more efficiently," Grooten explained.
The bimonthly VBSpam tests use Virus Bulletin's live email feed as well as a spam messages provided by Project Honey Pot. Each tested product is exposed to the same email stream.
More about the methodology (and the results) of the latest VBSpam tests can be found here. ®