Translation(s): English - Italiano


SpamAssassin is an extensible email filter which is used to identify spam. It has a wide range of features, uses DNSBL tests, heuristics, bayesian classification and other concepts to tell your spam from ham.

The official SpamAssassin homepage is at

Versions of Debian Packages

As SpamAssassin is a targetting a moving target, it is wise to keep it quite up-to-date, as spammers adapt to the filters to get their spew through.

Setting up Exim 4 and SA 2.64 to reject spam at SMTP time

<!> This section contains outdated information.

It is fairly easy to set up Exim4 and SA to reject spam in the SMTP dialogue, that is, while the spam is still in transmission and your system is still talking to the spammer.

First of all, you need the exim4-daemon-heavy which has exiscan-acl. The latter allows scanning messages. Then, you need SA installed, and just follow the documentation exiscan-acl-docs.


In the file /etc/exim4/conf.d/main/15_spamassassin-config I have simply the following:

  spamd_address = 783

To avoid scanning messages that are destined to postmaster or abuse (and is their damned duty to read) modify the entry in /etc/exim4/conf.d/acl/30_exim4-config_check_rcpt to read:

  accept local_parts = postmaster:abuse
         domains = +local_domains
         set acl_m0 = rfcnames

Then, the rest goes in /etc/exim4/conf.d/acl/40_exim4-config_check_data

Pretty high up, you need:

  accept condition = ${if eq{$acl_m0}{rfcnames} {1}{0}}

And then somewhere far down there:

  warn  message = X-Spam-Score: $spam_score ($spam_bar)
        spam = nobody:true

  warn  message = X-Spam-Flag: YES
        spam = nobody
  warn  message = X-Spam-Report: $spam_report
        spam = nobody

However, note that any errors can result in loss of valid e-mail, so make sure you have read and understand the official documentation before you use it. This is meant to give a taste of what is involved in getting it working.

Important Notes

It is controversial whether rejecting at SMTP time is such a good idea. The problem is, spammers (and viruses) routinely forge the from address on the envelope. This means that if there is a bounce generated, it will go to this address, which can be randomly generated, or worse, an innocent third party.

Therefore, it is very important that your system doesn't generate a bounce. That's your responsibility. For one thing, you can't accept a message and then bounce it, that would be wrong. You can't, for example, have your secondary MX accept a message and then have your primary reject it. If you are to use this, your primary and secondary MX should have identical configurations with respect to rejecting spam.

The other thing are relays. Spammers often take over relays to obscure the true source of the spams. If a relay under a spammer's control generate a bounce message based on your rejection, that bounce may go to an innocent thrid-party, which would be bad. That's the main reason why some consider this approach harmful.

The advocates of this approach points out that spammers have no interest in generating bounces, as that would only cut into their own spamming bandwidth and make them less efficient. Therefore, one rarely sees a raped relay, or a spammer himself, generate a bounce, which solves the problem.

Other advocates hold the position that "my system, my responsibility, your system, your responsiblity", implying that an administrator would need a heads-up he would get from letting his relay pass on spam.

FixMe: This page needs to be brought up to date.