Midrange Stuff, OS/400 Edition

The war against spam, like the war against terrorism, is an unwinnable war. The total eradication of unsolicited e-mail, no matter how desirable it may seem, is an untenable goal. In lieu of absolute victory, anti-spam forces focus on containing spam and outwitting the spammers, who continually adapt their techniques to sneak through spam defenses. This is the approach taken by GROUP Technologies, which recently equipped its anti-spam software with new content-recognition technology that, the company says, adapts to new forms of spam better than prevailing methods.

As the amount of spam e-mail has increased over the years, security software companies have employed a variety of methods to root out spam and delete it, with varying degrees of success. They've used sophisticated algorithms to create spam filters and have compiled lists of spam keywords and known spammers. But no matter how resourceful the anti-spam forces, spammers have found creative ways around the defenses, says Jacquelyn Thrasivoulos, a GROUP Technologies spokesperson. "Spammers have become more creative as they learn how anti-spam technology works, and what types of black lists and white lists there are, and the top 50 words on the banned list," she says. "In terms of the game, it's a very costly one."

Indeed, spam is one area of the economy that is actually growing. The amount of spam doubled from about 10 percent in late 2001 to more than 20 percent by the end of 2002. Now, only four months into 2003, the amount of spam e-mail has doubled again, and accounts for 45 percent of e-mail traffic, according to Brightmail, an anti-spam software provider. All that spam will cost about $10 billion in 2003 in lost worker productivity at American corporations, according to Ferris Research, a market research firm.

To identify spam for destruction, most anti-spam products use a filter that is powered by an algorithm that provides some degree of adaptability, or learning capability. GROUP Technologies had certain characteristics in mind as it set out for a new algorithm to power the spam filter in its securiQ.Wall spam-killing software. "We were looking for an algorithm with very fast processing and self-learning capabilities," says Thrasivoulos. "We found a technology that was superior to the Bayesian algorithm, which a lot of the vendors are using."

GROUP Technologies asked Guild Companies not to identify the name of the algorithm it chose to use, except to say that it came out of a European artificial-intelligence university project in the early 1990s and has been in widespread use in Europe for about 10 years. Needless to say, GROUP, headquartered in Germany, considers its choice of algorithm, which powers its new Content Recognition Engine, or CORE spam filter, a competitive advantage in its continuing campaign against spam. "We've chosen not to go with the Bayesian algorithm, which differentiates us," Thrasivoulos says.

Other enhancements that GROUP has rolled out with its new CORE filter in securiQ.Wall include support for more types of e-mail attachments, including PDFs and zipped, encrypted, and compressed files. While external spam is not often delivered in these formats, internal spam is. "When people think of spam, they think of it coming into the organization," Thrasivoulos says. "However, there's a greater risk of information being disseminated within an organization that's of a non-business function" than of spam originating from the outside.

For example, employees who e-mail a joke, a music video, or a pornographic image to others from their PC at work are just as guilty of sending spam as the obnoxious money lenders that proliferate on the Internet. And when the attachment is big, as with a video file, it becomes as much a bandwidth issue as it does a productivity issue. Today, securiQ.Wall supports more than 70 different types of file formats that are commonly attached to e-mails.

The new version of securiQ.Wall begins shipping next month. securiQ.Wall supports IBM Lotus Domino and Microsoft Exchange e-mail servers, running on a variety of platforms, including OS/400, AIX, Linux, Solaris, and Windows. An annual subscription to securiQ.Wall starts at $3,162 for 150 users, or about $21 per seat. For more information, go to www.group-technologies.com.