Configuring Bayesian options
The Bayesian section of antispam profiles lets you configure the FortiMail unit to use Bayesian databases to determine if the email is likely to be spam. If the Bayesian scan indicates that the email is likely to be spam, the FortiMail unit treats the email as spam and performs the associated action.
FortiMail units can maintain two Bayesian databases: global and per-domain.
• For outgoing email, the FortiMail unit uses the global Bayesian database.
• For incoming email, which database will be used when performing the Bayesian scan varies by configuration of the incoming antispam profile and the configuration of the protected domain.
Before using Bayesian scans, you must train one or more Bayesian databases in order to teach the FortiMail unit which words indicate probable spam. If a Bayesian database is not sufficiently trained, it can increase false positive and/or false negative rates. You can train the Bayesian databases of your FortiMail unit in several ways. For more information, see
“Training the Bayesian databases”.
| 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. |
To configure Bayesian scan options
1. When configuring an antispam profile, enable Bayesian in the AntiSpam Profile dialog.
2. Click the arrow to expand Bayesian.
3. From Action, select the action profile that you want the FortiMail unit to use if the Bayesian scan finds spam email.
4. Configure the following:
GUI item | Description |
Accept training messages from users | Enable to accept training messages from email users. Training messages are email messages that email users forward to the email addresses of control accounts, such as is‑spam@example.com, in order to train or correct Bayesian databases. For information on Bayesian control account email addresses, see “Configuring the quarantine control accounts”. FortiMail units apply training messages to either the global or per-domain Bayesian database depending on your configuration of the protected domain to which the email user belongs. Disable to discard training messages. This option is available only if Direction is Incoming. (Per-domain Bayesian databases cannot be used when the recipient does not belong to a protected domain, which defines outgoing email.) |
Use other techniques for auto training | Enable to use scan results from FortiGuard, SURBL, and per-user and system-wide safe lists to train the Bayesian databases. This option is available only if Direction is Incoming. (Domain-level Bayesian databases cannot be used when the recipient does not belong to a protected domain, which defines outgoing email.) |
5. Continue to the next section, or click Create or OK to save the antispam profile.