Question How to reject SPAM instead of Delete, Move or just Mark

[hansitheking]

Many of my Users like to reject SPAM messages. Plesk SpamAssassin is only offering delete, move to spam folder or just mark in subject.

My solution was to mark spam messages in the subject line and I told my Users to create a sieve rule to reject the email, the problem is that the reject message is not getting send out and in the mail log i can find:

Sep  2 10:47:30 server dovecot: service=lda, user=user@example.com, ip=[]. sieve: msgid=<2347892-4589-5&f-8F73-8210F0E0AF85@from-example.com>: not sending reject message to <>

Why is the reject messages not sent?
Does any one of you have another solution to reject SPAM back to the sender?

 

[mr-wolf]

ASSP is a very good Anti-Spam reverse SMTP proxy.
It comes between the Internet and your Postfix.
I use an iptables nat rule to insert the proxy between the outside world and postfix. This way you don't need to change Plesk parameters.
It is NOT a transparent proxy and it therefore needs some real knowledge of SMTP.

I've been using it more than 10 years.
I can't recommend it to people that like to install that with some clicks, though.

I posted some stuff here, but it takes more than that...
Resolved - ASSP integration with plesk postfix

 

[G J Piper]

You can set up a (Dovecot sieve) filter in Roundcube that reads the Spamassassin email headers and does any number of things with it — including reject with a message, delete, move, etc. This is a per-user solution though not really a global one.

For a global solution you would need a global sieve. Here are instructions for how to create a simple spam deletion sieve (outside the GUI) I created:
Resolved - Global Dovecot Sieve Rule?

Not sure it is ever a good idea to "reject with a message" announcing your email is valid to all spammers, however. You may get sending errors like the one you are experiencing due to spam message sending servers being "spoofed" and therefore unable to receive your reject message.

 

[mow]

You can set up a (Dovecot sieve) filter in Roundcube that reads the Spamassassin email headers and does any number of things with it — including reject with a message, delete, move, etc.

The problem with that approach is, what the above describes is rather a bounce than a reject.
What you want when you deal with spam is to reject it directly in the SMTP dialog (like it is done for greylisting) because the SMTP connection is the only thing you know for sure about the sender; everything else can, and will, be faked. For this to work, Spamassassin must be hooked into postfix (like postgrey is) so it can process the message's header and content immediately after they were received and tell postfix whether to respond to the message content with "250 Message accepted" or "550 do not want".
Dovecot/Roundcube only gets the message after it was already accepted (and the connection closed) by postfix, and thus, it cannot inform the sender about the rejection directly over a still-open SMTP connection and would have to rely on the data supplied by the alleged spammer to create a bounce - which would make it possible to use the server as a spam reflector.

 

[hansitheking]

@mow Thank you very much for your very good explanation! Do you have any hint for my how to hook SpamAssassin into postfix?