2007-12-10

Thunderbird and everyday spam fighting

How many emails do you get every day? How many of them are not spam?
I get about 200 spam emails everyday. I'm using Thunderbird mail client with built in spam filter - but it "eats" only about 15% of my spam. I think that it's not the matter of training data because of about several thousands good marked posts.

Somebody wrote about manual preselecting best representatives for the types of spam, and then feeding the spam filter with it. But come on - I need and automated method.

So I started looking for the best spam filters I can use for free. Very popular and tempting option is to forward all your emails to Google mail and use it as your mail client. Filtering capabilities are indeed very good, but what if you don't want to redirect all your email to "G empire"?

There are good open source spam filters working as proxy servers like fabulous spamassasin. It's good idea to set up such kind of spam filtering server for intranet.

I've needed something different. While searching for Thunderbird spam filtering plugins I found adn tried one interesting product - Spamato.

It can work as standalone mail proxy but I used Thunderbird plugin version spamato4thunderbird. There is also MS Outlook version too if anybody is interested.

Installation procedure is easy - like for most of hunderbird plugins. Basic configuration - show your bin/java path and give registration email (spamato can send spam statistics to central server). Oh I mentioned java. Yes it's written in java- and proces takes about 70 MB under windows, and it's not blazing fast. It's not a problem for me - I'm checking email a couple times per day and don't need to run Thunderbird all day. It's the matter of work organization.

What you get after install: a beast with pluggable architecture, a dozen filter and decision maker modules, configurable via http interface, and displaying nice charts. Oh and there is that funny (annoying) sound after every spam email marked - it's good you can easy turn it off.

Using experience- surprisingly good. Filtering results - not worse than Google mail filter. There were few false positives at the beginning (ham marked as spam), but all those filters are learning. All spam marked messages are moved into special spamato folder (check configuration). You can view spam history and correct (teach) filters via web interface. Big negative - spamato is not integrated with Thunderbird native junk mail tools. So marking something as junk under Thunderbird, doesn't count for Spamato. You have to correct filters on Spamato separately. There is partial solution for that problem - you can configure spamato to detect posts movements in/out of spamato folder on IMAP server.

So I'm going back to read my spam errrgh email inbox.

No comments: