Distributed James Server — Anti-Spam configuration

Anti-Spam system can be configured via two main different mechanisms:

  • SMTP Hooks;

  • Mailets;

AntiSpam SMTP Hooks

"FastFail" SMTP Hooks acts to reject before spooling on the SMTP level. SpamAssasin hook can be used as a fastfail hook, therefore SpamAssassin must run as a server on the same machine as the Apache James Server.

SMTP Hooks for non-existent users, DSN filter, domains with invalid MX record, can also be configured.

SpamAssassinHandler (experimental) also enables to classify the messages as spam or not with an configurable score threshold (0.0, non configurable). Only a global database is supported. Per user spam detection is not supported by this hook.

AntiSpam Mailets

  • SpamAssassin Mailet is designed to classify the messages as spam or not with an configurable score threshold. Usually a message will only be considered as spam if it matches multiple criteria; matching just a single test will not usually be enough to reach the threshold. Note that this mailet is executed on a per-user basis.

Here is an example of mailet pipeline conducting out SpamAssassin execution:

<mailet match="All" class="SpamAssassin">
    <onMailetException>ignore</onMailetException>
    <spamdHost>spamassassin</spamdHost>
    <spamdPort>783</spamdPort>
</mailet>
<mailet match="All" class="MailAttributesToMimeHeaders">
    <!-- This mailet is not required, but useful to have SpamAssassin score in headers-->
    <simplemapping>org.apache.james.spamassassin.status; X-JAMES-SPAMASSASSIN-STATUS</simplemapping>
    <simplemapping>org.apache.james.spamassassin.flag; X-JAMES-SPAMASSASSIN-FLAG</simplemapping>
</mailet>
<mailet match="IsMarkedAsSpam" class="WithStorageDirective">
    <targetFolderName>Spam</targetFolderName>
</mailet>

  • BayesianAnalysis (unsupported) in the Mailet uses Bayesian probability to classify mail as spam or not spam. It relies on the training data coming from the users’ judgment. Users need to manually judge as spam and send to spam@thisdomain.com, oppositely, if not spam they then send to not.spam@thisdomain.com. BayesianAnalysisfeeder learns from this training dataset, and build predictive models based on Bayesian probability. There will be a certain table for maintaining the frequency of Corpus for keywords in the database. Every 10 mins a thread in the BayesianAnalysis will check and update the table. Also, the correct approach is to send the original spam or non-spam as an attachment to another message sent to the feeder in order to avoid bias from the current sender’s email header.

Feedback for SpamAssassin

If enabled, the SpamAssassinListener will asynchronously report users mails moved to the Spam mailbox as Spam, and other mails as Ham, effectively populating the user database for per user spam detection. This enables a per-user Spam categorization to be conducted out by the SpamAssassin mailet, the SpamAssassin hook being unaffected.

The SpamAssassin listener requires an extra configuration file spamassassin.properties to configure SpamAssassin connection (optional):

Table 1. spamassassin.properties content
Property name explanation

spamassassin.host

Hostname of the SpamAssassin server. Defaults to 127.0.0.1.

spamassassin.port

Port of the SpamAssassin server. Defaults to 783.

Note that this configuration file only affects the listener, and not the hook or mailet.

The SpamAssassin listener needs to explicitly be registered with listeners.xml.

Example:

<listeners>
  <listener>
    <class>org.apache.james.mailbox.spamassassin.SpamAssassinListener</class>
    <async>true</async>
  </listener>
</listeners>