SecurityGateway and MDaemon both feature Bayesian learning, which allows administrators (or users, when authorized) to feed samples of spam and non-spam email messages to designated public folders. By default, when 200 samples of spam and 200 samples of non-spam have been placed in these folders, the Bayesian learning process will process these folders and feed their contents to a database of what are known as tokens – snippets of spam-like and ham-like (non-spam) content, basically. We all know that we humans are not infallible – people make mistakes, so it’s possible for messages to be fed to the wrong folders. When this happens, users may begin to receive more false-negatives (spam that was not caught by the spam filter) or you may accumulate a number of false positives (legitimate email messages that were flagged as spam by the spam filter). When this happens, it may be necessary to rebuild the Bayesian database. You may recall that I posted a blog entry awhile back on how to rebuild the Bayesian database for MDaemon. You can read that post here. For SecurityGateway, the concepts are the same, but the navigation and file locations are different. The following tutorial video explains how to rebuild the Bayesian database in SecurityGateway.
WorldClient, MDaemon’s webmail client, features icons that users can click on to feed examples of spam or non-spam to MDaemon’s Bayesian learning feature.
MDaemon’s Bayesian Learning feature allows you to feed samples of spam and non-spam to designated folders. These samples are used to train the spam filter to recognize spam and non-spam messages. The more samples you feed these folders, the more accurate your spam filter can be over time. This feature is especially useful if you work in an industry that uses medical or financial terms. If you work for a bank, messages containing pharmaceutical terminology can be handled with greater scrutiny, while messages of a financial nature can be treated with less suspicion. The same is true if you work for a medical facility. Your daily email correspondence may contain messages with names of medications or other pharmacy related terms, while they may not contain many words of a financial nature. So if you’re a bank & you receive a great deal of spam with the names of medications in them, you can feed those messages to your Bayesian Learning folder so that MDaemon can learn from them based on your specific business.
Using WorldClient, MDaemon’s webmail client, users can feed examples of spam and non-spam to MDaemon’s Bayesian Learning system using the Thumbs Up & Thumbs Down icons in the upper toolbar. But what if these icons are missing for a particular user, or for all users? Or, what if you as the administrator do not want users to have access to the Bayesian learning folders due to messages being improperly fed to the wrong folders? In this situation, you could create two more folders for spam and non-spam messages, then grant your users access those folders instead of to the main Bayesian learning folders. This means, though, that the thumbs up & thumbs down icons will not appear in WorldClient, but users would still be able to access the new public folders you created (if you granted them access via Public Folder Manager) for placement of samples of spam or non-spam.
I’ve put together the following tutorial video, which shows how to enable the Thumbs up & Thumbs down icons in WorldClient.