Roaring Penguin adds Bayesian filters to anti-spam software

Software maker Roaring Penguin Tuesday released an upgrade to its CanIt anti-spam program that includes Bayesian filters to help the software recognize spam and block it before it reaches end users’ mailboxes.

CanIt 2.0, available immediately, “learns” what an organization deems spam and what is considered legitimate e-mail through Bayesian analysis, a technique based on probabilities, says Roaring Penguin president David Skoll. With CanIt Pro 2.0, a version of the anti-spam software that lets a company’s end users set controls and preferences, each employee can train their own Bayesian filter to block spam based on his or her needs.

“The biggest thing with the new version is the Bayesian filtering, which is minimizing my administrative overhead,” says John Mason, information systems manager with Autostrade International of Virginia OAM, which manages a private toll road near Dulles airport. Mason is beta testing CanIt 2.0 and says the software has learned, for example, that e-mail messages with very little content in it other than a URL are spam, he says.

Autostrade International recently switched from CanIt to CanIt Pro so end users could have more control over the software, Mason says. The company also switched to the professional version because “I knew the product’s development path was going towards Bayesian filtering, and that’s more effective at a per-user level,” he says.

Version 2.0 includes a technique Skoll calls “hit and run,” also known as “gray listing,” that detects spam by forcing the sender to resend their message. The first time a user of CanIt 2.0 receives mail from a new sender, the program automatically sends a reply telling the sender there was a temporary failure and asking them to resend the message. “Most spammers’ software doesn’t bother retrying,” says Skoll, and so the unwanted message never reaches the end user. As for legitimate messages, “the worst thing that can happen is there’s a 15- to 30-minute delay” in receiving the e-mail, he says.

The upgrade to CanIt also features a simplified user interface, so novice users can work with a simple set of menu options while more advanced users can click a button to access features such as white and black lists, Skoll says. Roaring Penguin also provides access to CanIt’s source code under a license agreement, so organizations can modify the program to fit their use, Skoll says.

CanIt can be installed on an organization’s mail server or at the Internet gateway. CanIt 2.0 is priced starting at $6 per inbox per year for the first year, with a 50% discount for subsequent years. Volume discounts apply; for example, an organization with 5,000 inboxes would pay $3.15 per inbox for the first year.