Training Bayesian database

[#1]

Article

Bayesian scanning is one of the key technologies the FortiMail unit uses to filter email for spam. The FortiMail unit uses an account system to train the Bayesian databases that are the core of Bayesian scanning. For details about Bayesian database training, see the FortiMail Administration Guide.

By training the Bayesian databases, email users can help to improve the accuracy of the database by selectively forwarding messages to the FortiMail unit. A database is said to be well-trained when it becomes more accurate at catching spam through this method.

Training the Bayesian databases involves two steps:

1. First, email users must forward spam and non-spam messages to the FortiMail unit (see the first two bulleted items in the box below). This is especially important when your databases are empty.
2. Later, email users can forward spam the FortiMail unit has failed to catch, or email the FortiMail unit has incorrectly detected as spam, to the FortiMail unit to fine-tune the databases (see the third and fourth bulleted items in the box below).

The FortiMail system administrator is responsible for setting up the FortiMail Bayesian accounts for email users to send spam information to. The administrator also needs to send email users instructions similar to the following on how to train the Bayesian databases:

To train the Bayesian databases:

• If you have collected spam email messages and want to train your personal Bayesian database on the FortiMail unit, forward them to learn-is-spam@example.com from your company email account. This ensures that similar email will be tagged as spam by the FortiMail unit in the future.

• If you have collected non-spam email messages and want to train your personal Bayesian database on the FortiMail unit, forward them to learn-is-not-spam@example.com from your company email account. This ensures that similar email will not be tagged as spam by the FortiMail unit in the future.

• If you receive spam email messages that have not been caught and tagged by the FortiMail unit, forward them to is-spam@example.com from your company email account to ensure that similar email will be caught by the FortiMail unit in the future.

• If you receive email messages that the FortiMail unit has incorrectly tagged as spam, forward them to is-not-spam@example.com from your company email account to ensure that similar email will not be tagged as spam by the FortiMail unit in the future.

• If you belong to an email alias and receive a spam message sent to the alias address, forward it to the FortiMail "is-spam" Bayesian account to train the database of the alias address. Remember to enter the alias address, instead of your own, in the "From" field.