[SWLUG] help required with spamassassin

Jonathan Wright mail at djnauk.co.uk
Sun Jul 30 10:06:09 UTC 2006


Andy Dixon wrote:
> Hello,
> 
> If you can move any spam messages you get into a mailspool file, you can 
> use sa-learn to learn the spam messages.
> 
> http://spamassassin.apache.org/full/3.0.x/dist/doc/sa-learn.html
> 
> If afterwards you then run through all your legit mail as ham, that will 
> help to eliminate false positives.

To use this effectively, you'll need at least 1000 of each. But, I used 
to run sa-learn regularly, and over time is does really try to eat some 
memory. Esp as you cannot run it as a single daemon (it must split into 
at least two), and each has it's own copy of the data (doesn't share).

Personally, I just run DCC and RulesDuJour as additional options for 
SpamAssassin any my detection rate is around 90-95%.

DCC (Distributed Checksum Clearinghouse) calculated the checksum of the 
message (like MD5 it) and checks with the server if it's know Spam. 
Doesn't send the actual e-mail, but has an advantage in that it's very 
dynmaic. If alot of hosts start reporting the same checksum very 
quickly, it can be reported as Spam.

RulesDuJour is a bash script which downloads selected 'extra' rules, 
which can cover things such as words (including other languages), sites 
know to be used in Spam e-mails (although that's very large and you'll 
need about 1Gb of RAM to really use that one).

You just run it every night and it'll download updated files and restart 
SpamAssassin for you.

-- 
  Jonathan Wright
   ~ mail at djnauk.co.uk
   ~ www.djnauk.co.uk
--
  2.6.17-gentoo-r3-djnauk-b1 AMD Athlon(tm) XP 2100+
  up 9 days, 10:35, 1 user, load average: 3.60, 2.23, 1.50
--
  cat /dev/random (because u never know, u may see something u like)
--
  "Homosexuality is God`s way  of  ensuring  that  the  truly  gifted
  aren't burdened with children."

                                                         ~ Sam Austin



More information about the Swlug mailing list