Air-Stream Mail Services now utilise the DSPAM content filtering package to filter incoming emails.
DSPAM is another fine open source product, see the SourceForge project page for more info: http://dspam.sourceforge.net.
How it Works
As emails arrive they are identified as spam or not spam and tagged with a special signature so that the system can remember its choice. You can see this signature along with some other info if you view the message headers:
X-DSPAM-Result: Innocent X-DSPAM-Processed: Tue Aug 18 18:21:18 2009 X-DSPAM-Confidence: 0.9899 X-DSPAM-Probability: 0.0000 X-DSPAM-Signature: 4a8a6b86266376863418247
Since the DSPAM starts out knowing nothing, initially it thinks nothing is spam, so users are required to train the filter by telling it what is and is not spam. This is done by either moving the falsely identified message to a special folder, or forwarding them to the DSPAM system for retraining. DSPAM then reads the signature from the email and uses it to change its decision about that message.
After a bit of training, DSPAM will be able to accurately recognise spam messages. To ensure it doesn't filter out any of your legitimate mail you should check the SPAM folder regularly, read on to find out how.
Checking the SPAM Filter
When a message arrives that is identified as spam it will prefix the subject line with the word [SPAM] and moved it into a special "SPAM" folder within your Inbox. To check what's in this folder simply log onto the Web Mail or have a look in your IMAP client.
Use the training methods below if you find DSPAM is incorrectly tagging your messages.
Training the Filter: Method 1
As with any learning content filter, occasionally the system gets it wrong and either doesn't tag a message a spam when it really is (false-negative), or tags it as SPAM when it really isn't (false-positive).
For those of you who use IMAP or Web Mail, training the filter is simple. A new namespace called "Public" has been created, with 2 folders HAMLEARN and SPAMLEARN. Users can move messages into these folders and they will be used to retrain the system periodically, then removed. These folders are visible by all users so keep in mind that anything you move there can be seen by others.
SPAMLEARN: Move false-negatives into this folder. If a spam makes it into your inbox, you should put it here.
HAMLEARN: Move false-positives into this folder. If a legitimate message is filtered out of your inbox (you found it in the SPAM folder), you should put it here (as long as you don't mind everyone being able to read it). Alternatively see Method 2 below.
Training the Filter: Method 2
As an alternative to the above method, you can also forward any falsely identified emails to the following addresses:
ham _at_ air-stream dot org: For false-positives.
spam _at_ air-stream dot org: For false-negatives.
If you use this method, ensure that the email is being forwarded in its entirety, with all email headers included.
