Training Bayesian databases
Bayesian scanning can be used by antispam profiles to filter email for spam. In order to be accurate, the Bayesian databases that are at the core of this scan must be trained. This is especially important when the databases are empty.
Be aware that, without ongoing training, Bayesian scanning will become significantly less effective over time and thus Fortinet does not recommend enabling the Bayesian scanning feature.
Administrators can provide initial training. For details, see Training the Bayesian databases. If you have enabled it (see Configuring the Bayesian training control accounts and Accept training messages from user),email users can also contribute to training the Bayesian databases.
To help to improve the accuracy of the database, email users selectively forward email to the FortiMail unit. These email are used as models of what is or is not spam. When it has seen enough examples to become more accurate at catching spam, a Bayesian database is said to be well-trained.
For example, if the local domain is example.com, and the Bayesian control email addresses are the default ones, an administrator might provide the following instructions to his or her email users.
To train your Bayesian filters
- Initially, forward a sample set of spam and non-spam messages.
- If you have collected spam, such as in a junk mail folder, and want to train your personal antispam filters, forward them to
learn-is-spam@example.comfrom your email account. Similar email will be recognized as spam. - If you have collected non-spam email, such as your inbox or archives, and want to train your personal spam filters, forward them to
learn-is-not-spam@example.comfrom your email account. Similar email will be recognized as legitimate email.
- Forward undetected spam to
is-spam@example.comfrom your email account. - Forward legitimate email that was mistaken for spam to
is-not-spam@example.comfrom your email account. - If you belong to an alias and receive spam that was sent to the alias address, forward it to
is-spam@example.comto train the alias’s database. Remember to enter the alias, instead of your own email address, in theFrom:field.
This helps your antispam filters to properly distinguish similar email/spam in the future.