I personally just have a script parse my spam folder each night and
send them to the learner.
I also can't locate any meaningful info in those emails to id them with.
The magic jack is just one of like 20-30 different variations of that
spam source.
They are just hacking webservers, and sending the spam from them, and
looks like they move webservers that they hacked every hour or two.
Just making blacklisting or tracking it basically impossible.
I did setup my own blacklist, that works kind of like the
mailspike.org Z list, that when a spam wave hits, they get blacklisted
for 5hours, that seems to be working surpisingly well. Not sure
exactly why the mailspike one isn't working as good. Maybe cause they
are only basing it on spamtraps? and I'm doing it just based on spam
average per ip in the last hour.
Quoting Benny Pedersen <[email protected]>:
Den 2012-02-26 01:53, Elijah Savage skrev:
I was hoping someone would recommend which scores they may adjust.
that would be incorrect to adjust scores, better is to create tests
in spamassassin to hit on this specific spams, simplest way is
sa-learn --spam msgfile
i can see you using dspam with amavis ?
how do you train it ?
i would like to use dspam in amavis, but so far found
dovecot-antispam works with dspam
