Help & How-to

Bayes setting on Roundcube


ServerMX makes several webmail clients available to its customers. In Roundcube you can configure SpamAssassin, our antispam, from Settings → Junk; among the options there are the Bayes settings.

Bayes settings improve the control over spam. The name comes from the "Bayes Theorem" of probability: the Bayesian classifier learns to identify spam from the messages that were classified as spam (or not spam) in the past.

We advise changing the default Bayes configuration carefully: these options are meant for advanced users.

The following figure shows the Bayes options in Roundcube (Settings → Junk → Bayes Settings).

Roundcube, Junk, Bayes Settings

 

1) Main options - Use Bayesian style classifier

Turn it on to let SpamAssassin classify messages as "Junk" or "Non junk" with the Bayesian classifier.

2) Main options - Use Bayesian rules

Click the information icon next to each option to read the help provided by the plug-in:

Bayes configuration - Use Bayesian rules

 

3) Main options - Use Bayes auto learn

With auto learn the system automatically uses the messages that are very likely junk or non junk to improve detection:

Bayes configuration - Use Bayes auto learn

 

4) Auto learn options - Non junk threshold

SpamAssassin "learns" what non junk messages look like from the messages with a spam score below this value, 0.1 by default (the spam score is shown in two formats: decimal or integer, so 0.1 in decimal format is 1 in integer format).

5) Auto learn options - Junk threshold

SpamAssassin "learns" what junk looks like from the messages with a spam score above this value, 12 by default (120 in integer format).

Bayes configuration - Non junk and Junk thresholds

 

Click Save to store your options.

Clicking Restore Default Settings restores the default settings. Please keep in mind that the Bayes plug-in works together with the Address Rules: if you confirm "Restore Default Settings" you also delete your previous rules, including your address rules.

In conclusion, we suggest that only advanced users change the Bayes settings, and that they monitor how spam detection changes afterwards. Configuring this plug-in is an iterative process: check the effect of each change to find the right configuration for your needs.

Our contacts