SpamAssassin Maintenance and Plugin Development

One of the areas that I work in most frequently is spam filtering, and a lot of my professional life has been devoted to not just setting up and maintaining filtering systems, but also increasing their accuracy by developing spamassassin plugins and feedback mechanisms

As well as writing custom rules, I've authored several custom plugins for clients (sadly none were made public - either similar plugins existed, or they were too client-specific to be suitable for widespread use). These include:

A more flexible OCR (Optical Character Recognition - for detecting spam text in images) plugin, which took into account image type, dimensions, and had a more flexible approach to scoring.
Detection of the sender's operating system using p0f (passive fingerprinting). This allowed us to penalise Windows Vista/XP/9x (since the majority of mail servers use either some flavour of UNIX/Linux, or server-orientated version of Windows)
Scanning of MS Word documents, PDFs etc for spammy content
Detection and filtering of so called scatter back (a spammer sends mail with your address as the sender, you receive all the bounce messages)

Mail servers come in all shapes and size, but I have experience in the following setups:

Postfix + SpamAssassin
Exim + SpamAssassin
Qmail + SpamAssasin (via Qmail-Scanner and .qmail) + Dspam
Sendmail + MIMEDefang + SpamAssassin

The latter was a heavily modified setup, with a web front-end which allowed users to control most aspects of their filtering (whether to use Bayes, which DNSBLs to use, required score, whitelisting, etc).

Of particular interest was a series of modifications I applied to Plesk to allow 1) mail over a certain threshold to be moved to a separate mail directory, and 2) allow webmail users to feed mail to the Bayes database via 'report as spam' style buttons.

If you have a mail server that you'd like me to take a look, feel free to get in touch; I've yet to come across a server on which I was unable to make a significant impact on the amount of spam being received by end users.

Services

Code

vBulletin

Data

Fun Stuff

pete@linuxbox.co.uk