Welcome to SpamBayes

SpamBayes is an Outlook plugin that provides a spam filter based on statistical analysis of your personal mail. Unlike many other spam detection systems, SpamBayes actually learns what you consider spam, and continually adapts as both your regular email and spam patterns change.

When you first start Outlook after SpamBayes has been installed, the SpamBayes Installation Wizard will appear. This Wizard will guide you through the configuration process and allow you to quickly have SpamBayes filtering your mail. This document contains additional information which will help make SpamBayes effective from the first time you use it.

Please remember that this is free software; please be patient, and note that there are plenty of things to improve. There are ways you can help, even if you aren't a programmer - you could help with this documentation, or make a donation, or any number of other ways - please see the main project page for information. The list of Frequently Asked Questions may also answer some of yours.

For more information on SpamBayes, including links to the rest of the documentation, see About SpamBayes.

Training

SpamBayes has no builtin rules, so anything you consider spam will be treated as spam by this system, even if it does not conform to the traditional definitions of spam. This means that SpamBayes requires training before it will be effective. There are two ways that this training can be done:

Allow SpamBayes to learn as it goes. Initially, SpamBayes will consider all mail items unsure, and each item will be used to train. In this scenario, SpamBayes will take some time to become effective. It will rarely make mistakes, but will continue to be unsure about items until the training information grows.
Pre-sorting mail into two folders; one containing only examples of good messages, and another containing only examples of spam. SpamBayes will then process all these messages gathering the clues it uses to filter mail. Depending on how many messages you train on, SpamBayes will be immediately effective at correctly classifying your mail.

The Installation Wizard will guide you through these options.

It is important to note that even when SpamBayes has little training information, it rarely gets things wrong - the worst it generally does is to classify a message as unsure. However, as mentioned, the more training information SpamBayes has, the less it is unsure about new messages. See using the plugin below for more information.

Using the Plugin

This section describes how the plugin operates once it is configured. You can access the SpamBayes features from the SpamBayes toolbar (shown to the right), but in general, SpamBayes will simply and silently filter your mail.

As messages arrive, they are given a spam score. This score is the measure of how "spammy" the system has decided a mail is, with 100% being certain spam, and 0% meaning the message is certainly not spam. The SpamBayes addin uses these scores to classify mail into one of three categories - certain spam, unsure, and good messages. Good messages are often known in the anti-spam community as ham.

Mail that is classified as good is never touched by this plugin - they will remain in your inbox, or be filtered normally by Outlook's builtin rules. Mail that is classified as either unsure or certain spam is moved into different folders for future review.

Outlook does not allow us to automatically add the spam score to your Outlook folder views - but you can do it manually by following these instructions.

There are three ways in which the system can get things wrong:

A spam stays in your inbox. This is known as a false negative. In this case you can either drag the message to the Spam folder or click on the Delete as Spam button on the Outlook toolbar. In both cases, the message will be trained as spam and will be moved to the spam folder.
Any message is moved to the unsure folder. In this case, the system is simply unsure about the message, and moves it to the possible spam folder for human review. All unsure messages should be manually classified; good messages can either be dragged back to the inbox, or have the Recover from Spam toolbar button clicked, while spam messages can either be dragged to the Spam folder or have the Delete as Spam toolbar button (shown above) clicked. In all cases, the system will automatically be trained appropriately.
A wanted (ham) message is moved to the Spam folder. This is known as a false positive. In this case you can either drag the message back to the folder from which it came (generally the inbox), or click on the Recover from Spam button. In both cases the message will be trained as good, and moved back to the original folder.

Note that in all cases, as you take corrective action on the mail, the system is also trained. This makes it less likely that another similar mail will be incorrectly classified in the future.