Setting Up BayesSpam

Calico Hosting's default junk mail protection is usually sufficient on its own. Popular email programs like Thunderbird provide another layer of junk protection too. But if you find that you often use the webmail interface to check email from remote locations, and you've already set Calico's Maia Mailguard to high protection, and some junk is still getting through, you can use BayesSpam Filtering to enhance your webmail experience.

Please read these instructions carefully!

Log into your webmail account and click on the Options link at the top. Scroll down and click on the BayesSpam - Intelligent Spam Filtering link.

To start using BayesSpam Intelligent Filtering, you must first build up a database of what you consider to be spam, and what you consider to not be spam. To do this, enable "BayesSpam links". When you read an email, there will be two new links to the upper right, to let you mark the mail as Spam or NonSpam. Use them a few dozen times on both spam and nonspam emails before going any further. The more times you use them, the more accurate your filtering will become. You'll also want to enable "Stats Tracking".

Once you have a few dozen trained emails in your database representing both spam and non-spam (check your DB stats on the BayesSpam settings page,) you can enable BayesSpam Filtering which will automatically filter out junk. If you wish, also set the 'Filtering Uncertain Messages' to Enabled. Make sure to set the corresponding filter boxes on the same page as well.

Now the filter will start deciding what is junk and what is not. It will leave good mail in the INBOX and filter Uncertain mail and mail deemed as junk to their respective folders/boxes. You need to be checking all three of those boxes (two if you are not using the 'Filter Uncertain Messages' option) to correct mistakes the filter makes. If it marks messages as junk that are not junk, go into the headers of the message and mark them as not junk and vice versa. Go back to Options-> BayesSpam - Intelligent Spam Filtering occasionally to look at the statistics of how accurate the filter is.

After you have done this with a few hundred messages and you are satisfied with the accuracy you might want to go back to the options and turn on Auto Learn. Again, you will need to track the three boxes for a little while to make sure the filter is not making mistakes.

After a while you should be able to leave the filter alone and just let it do its job. The amount of spam you see in your inbox should be minimal or none.

If something goes wacky in your spam training and you don't like the way things are being filtered, go to the options and click the 'Empty Database Now' link to start over from scratch. (This will usually not be necessary if you train the filter carefully.)

You have other options too. It is IMPORTANT that you understand what changing your settings will do. Please read on for a description of your other settings choices and suggested settings after you finish training the filter.

BayesSpam Probability Display - Adds a line to the header of each incoming message that gives the probability that the given message is spam. Enable this.

BayesSpam Filtering - Enable this only once you have manually set a few dozen messages to either Spam or Non-Spam.

Filtering Uncertain Messages - This option is only visible if you have enabled the BayesSpam Filtering option. When enabled, messages that the filter can't decide on will be sent to a separate mail folder. Enable this if you want.

Spam Deletion - Don't enable this. This option is somewhat dangerous, since if you get a false positive, it would delete the message without ever telling you. This option requires Filtering to be turned on as well.

Rebuild On Folder Refresh - Forces a database rebuild when folders are refreshed. Theoretically, unless you stay logged in all the time, disable this.

Rebuild On Login - This forces the filtering database to be rebuilt when you login. Enable this.

Auto Learn - This is the magic of the bayesian filter. Once you have trained it you can enable this option. Train the filter with hundreds of messages before turning this on. Auto Learn is still a little slow. Keep this in mind before enabling it. Auto Learn should not be enabled until you have manually added a number of spams AND nonspams to the plugin's database. Once it is doing a good job of filtering on it's own, then turn on AutoLearn, and it will automatically add every new email to it's database, thus improving it's filtering further.

Bytes of message to scan - This option limits how many bytes of a message the filter will scan. The bigger this number the longer it will take. Some messages can be quite large and would take a long time to scan. Leave at the default which is 65536 bytes. Make sure to hit the save button or your change will not be saved.

Folder To Filter Into - This is where messages that are decided to be spam are sent. Use a folder other than Trash as their may be some false positives (valid emails) that get sent here. Create a folder called JUNK and set this option to that folder. Make sure to hit the Save button or your change will not be saved.

Folder To Filter Uncertain Messages Into - This option is only visible if you have enabled both the BayesSpam Filtering and the Filtering Uncertain Messages options. This is the folder where messages would go if the Bayes filter is not able to decide if the message is spam or not. Do not use your trash box as there may be false positives (valid email) that go in here. Create a new folder and call it something like Uncertain or whatever else makes sense to you. Then set this option to that folder. Make sure to hit the Save button or this change will not be saved.

Check Inbox Only - If enabled only mail coming into your Inbox will be filtered. Messages you redirect to other folders by other means will not be filtered.

Add Folder To Ignore List - If a folder is added to the Ignore list messages coming into the folder will not be filtered. By default all folders are filtered. Make sure to hit the Save button or this change will not be saved. This is not 100% reliable, since BayesSpam may load before the main Filters plugin, thus giving it a chance to run on the messages in the inbox before they've been moved to their own folders.

Remove Folder From Ignore List - If a folder is removed from the Ignore list messages coming into the folder will be filtered. Make sure to hit the Save button or this change will not be saved.