Spam and other nuisance e-mail is a common problem for every Internet mail server. These messages clog inboxes, causing a loss of time and money. The Bayesian analysis feature in 602LAN SUITE 2004 provides an effective way to prevent users from ever seeing unwanted messages in their inbox.

Bayesian filtering

Named after the British mathematician Thomas Bayes, Bayesian filtering of Internet messages uses applied logic from Bayes’ theorem to determine the probability of a message that is spam (junk) or ham (not junk). The theorem provides a way to calculate
the occurrence of a target in future trials based on the occurrence of the target in past trials. When applied to e-mail, the Bayes’ theorem can be used to create a powerful anti-spam tool to calculate the probability that a received message is spam or ham based on previously received messages.

Bayesian Filtering in 602LAN SUITE

602LAN SUITE 2004’s Bayesian filter implementation can detect junk e-mail by actually looking at the e-mail content. This differs from the two other anti-spam methods provided in 602LAN SUITE, which blocks the originating IP address/server (DNSBL) or the originating e-mail address (blacklists), both of which can be easily circumvented by determined spammers. When a message is analyzed, a Bayesian score is added to the headers, which is displayed as “X-LNS-SpamCheck: NOSPAM, Bayesian=xx” for messages classified as ham (not junk) or “X-LNS-SpamCheck: SPAM, Bayesian=xx” for messages that are classified as spam (junk). The “xx” represents the probability score as determined by the Bayesian filter. This probability score is used by the Bayesian filter to trigger the “SPAM” or “NOSPAM” values. This can be adjusted on the Anti-spam
tab in the Advanced Configuration in 602LAN SUITE.

Training the Bayesian Filter

In order for Bayesian filtering to work properly in 602LAN SUITE, a minimum of 100 messages of both spam (junk) and ham (not junk) e-mail must be forwarded to the Bayesian filter for learning. Since the Bayesian filter uses past classification to determine the probability that a message is either spam or ham, sending additional messages of both types will yield better results. It is also recommended that the filter be reinforced with incorrectly classified messages. This will help the filter keep up with the ever-changing tactics spammers use to circumvent anti-spam filters.

More information

For more information on configuring and training the Bayesian Filter, please view the following links:

• Bayesian Anti-spam Configuration
• Eliminate Junk E-mail with 602LAN SUITE 2004’s Bayesian Anti-spam Filter!
• Configure 602LAN SUITE and Outlook Express to filter spam!