We use cookies on our website to ensure we provide you with the best experience on our website. By using our website, you agree to the use of cookies for analytics and personalized content.This website uses cookies. More Information
It seems like your browser didn't download the required fonts. Please revise your security settings and try again.

How should I train the Bayesian database on my Email Security Gateway?

  • Type: Knowledgebase
  • Date changed: 10 months ago
Solution #00001324

Scope:
This solution applies to all Email Security Gateways, firmware versions 3.3 and above.

Answer:
The Global Bayesian Database can be viewed under the Barracuda Bayesian Learning heading on the Basic > Spam Checking -> Bayesian/Intent section, and has the potential to affect messages addressed to all users when configured. It is trained via the web interface on the Basic > Message Log page by marking messages as Spam or Not Spam. The Per-User Bayesian Database is trained in the same fashion, either using the user Quarantine Inbox portion of the web interface or with an email client plugin (though these only affect mail for each individual user).

In order to be added to a Bayesian database, a particular message must exist in the Message Log (the entire body of the message must be present). If the Email Security Gateway blocked a message during the initial connection stages (for reason of RBL, invalid recipient, rate control, etc.) then the body of the message was never received, and it cannot be added to any Bayesian database. In addition, the message must also contain at least 10 words in the body of the message.

To train the global Bayesian database, follow these steps:
  1. Check the number of messages marked as Spam and Not Spam in the Bayesian database on the Basic > Spam Checking -> Bayesian/Intent section.
  2. Select an eligible message in the Message Log (on the Basic > Message Log page) by checking the box next to that message, and classify it as Spam or Not Spam by clicking the Spam or Not Spam button at the top of the page. For best results, mark false positives as Not Spam and spam getting through as Spam.
  3. Navigate back to the the Basic > Spam Checking ->  Bayesian/Intent section and verify that the Spam or Not Spam count has increased. If it has not, see Solution #00002289.
  4. Continue to mark messages as Spam and Not Spam until you have marked at least 200 messages each as Spam and Not Spam. You may mark more, but keep these numbers as close together as possible. Marking too many more as SpamNot Spam is not recommended; if your Bayesian Database is ineffective and you'd like to start over, reset it and start over instead of continuing to mark more messages as Spam and Not Spam.
    and
To train the Per-User Bayesian database for a user, follow these steps:
  1. Check the current count for that user's Bayesian database. To do this, from that user's Quarantine Inbox, navigate to Preferences > Spam Settings page.
  2. Select an eligible message in that user's Quarantine Inbox by checking the box next to that message, and classify it as Spam or Not Spam by clicking the Spam or Not Spam button at the top of the page. For best results, mark false positives as Not Spam and spam getting through as Spam.
  3. Alternatively, if that user has installed an email client plug-in, you may mark a message as Spam and Not Spam by clicking the Spam and Not Spam buttons in that user's mail client (after selecting that message).
  4. Go back to that user's Quarantine Inbox and navigate to Preferences > Spam Settings page to verify that the Spam or Not Spam count has increased. If it has not, see Solution #00002289.
  5. If a particular user's Bayesian Database contains at least 200 Spam messages at least 200 other Not Spam messages, that user's Per-User Bayesian Database will be used instead of the Global Bayesian Database for that user.
Additional Notes:
Spam, in recent years, has become more of a "moving target." Modern spam campaigns have become shorter and more targeted. As a result, your Email Security Gateway's Bayesian database may not be as effective as it once was unless it is constantly maintained by frequently adding similar numbers of Spam and Not Spam messages. Furthermore, with other recent feature enhancements, a Bayesian database is not crucial to effectively blocking spam messages to your email server.

Link to This Page:
https://campus.barracuda.com/solution/50160000000GQB5AAO