Spammers are animals but smart ones at that so they'll find a way around
your filters.
With spamassassin you must train against ham and spam for its bayesian
filter to work effectively.
In all honesty I never got spamassassin to tag spam to any of my
client's expectations. It was only after using Dspam that I saw the
results I wanted. Then, for no apparent reason started seeing influxes
of spam, with it as well, and had to retrain the database and eventually
move to another solution as my clients don't have time to retrain their
spam databases every six months.
The solutions I've started recommending are third party spam gateways.
I've got one client using the Sonicwall Anti-Spam gateway service which
works VERY well, although it doesn't have MTA to MTA TLS capability, and
another client that's going to go with there own Sonicwall Anti-Spam
appliance run as a VM, which does have MTA to MTA TLS capability. I know
of another individual who uses a Barracuda Anti-Spam appliance that
works well.
If you don't have time, or don't want, to deal with the headache of
fighting spam, and have the dough, I'd go third party. It saved my bacon.
If you can't go in that direction you might want to hop over to the
spamassassin and spamdyke mailing lists and see what kind of success you
can achieve with their help.
Hope that helps, Steve.
On 1/29/2017 4:09 PM, Steve Linberg wrote:
For a while, my Centos7 build was doing a pretty good job; once I
enabled Pyzor and blocked a number of suspect TLDs, I was getting spam
rejection rates averaging around 90%, but over the last month or so
it’s dropped to more like 65%.
I can’t see any obvious reasons for this; nothing unexpected in logs
as far as I can tell.
I’ve been saving spam that got through, and I’ve got a body of about
1500 spam messages now. I’m reading conflicting advice regarding the
use of sa-learn. If I run it against this body to teach it spam, do I
also need to do a “ham” run against OK mail? And are these numbers
high enough, or should I wait until I have more? Or is it risky to do
that at all?
Any advice for tuning this thing up? The spam problem seems to be
getting worse lately.
Thanks for any ideas.
--
Steve Linberg, Chief Goblin
Silicon Goblin Technologies
http://silicongoblin.com
Be kind. Remember, everyone you meet is fighting a hard battle.
--
Eric Broch, IMSO, DAM, NGOO, DITH, URTS
White Horse Technical Consulting (WHTC)
406.214.6802
