[Rspamd-Users] How to deliver spam anyway?

[Marjorie Roome]

Hi David,

On Sat, 2020-09-19 at 09:47 -0700, David Reagan wrote:
> I'd like to configure rspamd to deliver spam after it detects it. I
> have sieve configured to move spam into the Spam folder. That way
> users can check for any messages that got sent to spam when they
> shouldn't have.
> 
> I think I've configured /etc/rspamd/override.d/actions.conf to make 
> rspamd work that way, but I'm stuck on how to test..
> 
> My actions.conf:
> 
> ```
>  add_header = 6; # Add header when reaching this score
>  greylist   = 4; # Apply greylisting when reaching this score
> (will emit `soft reject action`)
>  subject    = "[SPAM] %s"
> ```
> 
> I'm hoping that making `reject = 15` not set will give me the
> behavior I want.
> 
> To test in my dev environment, I've set `enable_test_patterns =
> true;`, 
> and then I've been sending the GTUBE pattern. But as far as I can
> tell, rspamd just always rejects it. The only thing my override file
> did was let Thunderbird finishing sending the message. But the
> message never got past rspamd.
> 
> I see the logs in rspamd, but no logs show up in dovecot or postfix.
> 

Like yourself I run a family mailserver.

I have a largely similar set up (using sieve) but for rspamd actions
use:

Greylist       	4
Probably spam 	5.5
Rewrite subject [blank]
Spam 		15

This means that 'mild' spam ends up in my family's spam folders but
'strong' spam is rejected and not seen.

Initially I set Spam to 150 to see what I got and my experience is that
emails scoring more than 15 were never of interest. Also accepting all
spam only seems to encourage spammers. 
FWIW the GTUBE pattern will always get rejected whatever you set spam
to (even 150).

Greylisting at 4 means that 'mild' spammers get asked to send again,
this does avoid some processing if they fail to do so (some don't
retry).

Initially I used a Probably spam of 8, but once Bayes scoring kicked in
I was able to reduce it down to 5.5, which is the same as my threshold
when I used Spamassassin.

I avoid rewriting headers of 'mild spam' as if it is wrongly classified
and is then the recipient manually moves it back into their Inbox (and
triggers Bayesian learning) it will still have the '[SPAM']' header
there.

--
Marjorie