Trust The Octopus

In another post, Graham, one of the forum members.  Suggested that I put together a list of useful Open Source programmes that I recommend.

This has to be the first on my list.  POPFile

POPFile automatically sorts your messages and fights spam.
http://popfile.sourceforge.net/

POPFile won a Productivity Award at the 2005 Jolt Awards
In a short time, POPFile can achieve amazing accuracy - The Washington Post
It works for me - Kevin Polley

Basically you download POPFile.
Install it on your PC.
Set it up by following the instructions.

It sits between your email client (outlook etc) and your POP3 email account.
Before you get the email in your client POPFile classifies the messages, tossing them into buckets, according to your personal needs and requirements.

Basically you train it to your way of thinking.

For example:  You get lots of spam email offering you Viagra.  You teach POPFile by 'clicking a button' that this email is spam.
You do this for a week or two with every email you get.
Pretty soon POPFile starts to classify your spam as spam.

I send you an email which is a joke about Viagra but you still get to read it
Why?

Because the magic of POPFile has worked out that other emails 'I' sent you, you classified as maybe 'work' 'friend' etc and based on POPFiles logic, you'll want to read that email.

Basically it looks at every part of every email and breaks it down.  It assigns values to words, phrases, IP addresses, etc etc and then runs a calculation to decide what it thinks an email type is. 

If POPFile gets it wrong you correct it and it adds a 'balance' to further calculations based on your input.

It takes a while to be really efective but once it's trained it's REALLY good.
Here's my accuracy classification since Jan 4th 2004 to today.

Classification Accuracy
Messages classified:     150,981
Classification errors:     1,082
Accuracy:     99.28%
(Last Reset: Sun Jan 4 20:29:07 2004 )

AND .... it would be MUCH higher IF these figures only related to the actual *spam* mis-classifications.  These are below.
               False Positives    False Negatives
spam                293                      641

Many of the classification 'errors' I corrected - I didn't need to.  I was just being finicky about the 'type' of spam I was getting.

Being fair to POPFile - I'd say after processing 151,000 emails it's more like 99.8% accurate.

Go get it and try it yourself.
POPFile automatically sorts your messages and fights spam.
http://popfile.sourceforge.net/

Kev