Spam archives for training Spamassassin

Welcome!

We are fighting for 10 years against spammers, having very happy users. In these years we noticed that every second email passing through our server is spam, so we started to capture the spam emails in 2009, and now we have an every-day growing collection. Today we can tell that we have the biggest public collection on the Internet! Get it, and use it! Let's make the Internet a cleaner space!
These archives containing the spam emails captured within a 24 hours period are free for use according to GPLv2 licence! The format after unpacking with gunzip is mbox. You can use them for training spamassassin, dspam, bogofilter or other spam capturing software.

Do you have new-born system?

No problem! If you just installed Spamassassin and your Bayesian database is empty, you can use our startup Bayesian database, which is refreshed every day. Feel free to download it from here. After unpacking you will get a plain-text database created with sa-learn --backup > bayesian.database && gzip bayesian.database command.
The file size is 50 750 839 bytes (48.40 MB). If you are using Spamassassin, here it is a one-line command to download and init your own Bayesian database:

wget http://artinvoice.hu/spams/bayesian.database.gz && gunzip bayesian.database.gz && sa-learn --restore bayesian.database && rm -rf bayesian.database
Important! The command above will clear the existing Bayesian database, so it is not advised to use it on an already trained system. For training with more spams use the archives below.

If you would like to support this initiative, please help us by sending spams to spamreport@artinvoice.hu as .eml attachment.

Your donations are welcome to can maintain this server up and running!

If your system is already running for a while...

We have a solution for you as well. Use our daily captures to train you system. These emails are 100% spams and their source is the real life.
For automation you can use the following script:

#!/bin/bash
spamfile=spam--`date '+%Y-%m-%d'`.gz
spamfile_unpacked=spam--`date '+%Y-%m-%d'`
wget http://artinvoice.hu/spams/$spamfile
gunzip $spamfile
sa-learn --mbox --spam --progress $spamfile_unpacked
/etc/init.d/spamassassin condrestart
rm -rf $spamfile $spamfile_unpacked
exit 0


Note: If you want to train bogofilter you can use this command:

bogofilter -M -Ns -I $spamfile_unpacked

Important! To achieve the best results, train your filter regularly with ham (useful) emails as well! Ham and spam count should be nearly equal, but on a working system that will not be a problem. We cannot provide a ham collection because those are valid and good emails. If you run a mail server, you will have enough ham samples for training.