Emacs Gnus: Filter Spam

(draft for an upcoming book called Wicked Cool Emacs)

Ah, spam, the bane of our Internet lives. There is no completely
reliable way to automatically filter spam. Spam messages that slip
through the filters and perfectly legitimate messages that get
labelled spam are all part of the occupational hazards of using the
Internet.

The fastest way to filter spam is to use an external spam-filtering
program such as Spamassassin or Bogofilter, so your spam can be
filtered in the background and you don’t have to spend time in Emacs
filtering it yourself. In an ideal world, this would be done on the
mail server so that you don’t even need to download unwanted
messages. If your inbox isn’t full of ads for medicine or stocks, your
mail server is probably doing a decent job of filtering the mail for
you.

Server-based mail filtering

As spam filtering isn’t an exact science, you’ll want to find out how
you can check your spam folder for misclassified mail. If you download
your mail through POP, find out if there’s a webmail interface that
will allow you to check if any real mail has slipped into the junk
mail pile. If you’re on IMAP, your mail server might automatically
file spam messages in a different group. Here’s how to add the spam
group to your list of groups:

  1. Type M-x gnus to bring up the group buffer.
  2. Type ^ (gnus-group-enter-server-mode).
  3. Choose the nnimap: entry for your mail server and press RET (gnus-server-read-server).
  4. Find the spam or junk mail group if it exists.
  5. Type u (gnus-browse-unsubscribe-current-group) to toggle the subscription. Subscribed groups will appear in your M-x gnus screen if they contain at least one unread message.

Another alternative is to have all the mail (spam and non-spam)
delivered to your inbox, and then let Gnus be in charge of filing it
into your spam and non-spam groups. If other people manage your mail
server, ask them if you can have your mail processed by the spam
filter but still delivered to your inbox. If you’re administering your
own mail server, set up a spam filtering system such as SpamAssassin
or BogoFilter, then read the documentation of your spam filtering
system to find out how to process the mail.

Spam filtering systems typically add a header such as “X-Spam-Status”
or “X-Bogosity” to messages in order to indicate which messages are
spam or even how spammy they are. To check if your mail server tags
your messages as spam, open one of your messages in Gnus and type C-u
g (gnus-summary-show-article) to view the complete headers and
message. If you find a spam-related header such as X-Spam-Status, you
can use it to split your mail. Add the following to your ~/.gnus:

 (setq spam-use-regex-headers t) ;; (1)
 (setq spam-regex-headers-spam "^X-Spam-Status: Yes")   ;; (2)
 (require 'spam) ;; (3)
 (spam-initialize) ;; (4)

This configures spam.el to detect spam based on message
headers(1). Set spam-regex-headers-spam to a regular expression
matching the header your mail server uses to indicate spam.(2) This
configuration should be done before the spam.el library is loaded(3)
and initialized(4), because spam.el uses the spam-use-* variables to
determine which parts of the spam library to load.

In order to take advantage of this, you’ll also need to add a rule
that splits spam messages into a different group. If you haven’t set
up mail splitting yet, read qthe instructions on setting up fancy mail
splitting in “Project XXX: Organize mail into groups”. Add (:
spam-split) to either nnmail-split-fancy or nnimap-split-fancy,
depending on your configuration. For example, your ~/.gnus may look
like this:

(setq nnmail-split-fancy
'(
;; ... other split rules go here ...
(: spam-split)
;; ... other split rules go here ...
"mail.misc")) ; default mailbox

(draft for an upcoming book called Wicked Cool Emacs, more to come!)

  • Timo Lindfors

    Do you know an easy way to have gnus sort messages by their spam score efficiently? This way I would only need to “review” the messages that are just barely over the spam limit and could be real messages.

    • David

      That depends on the spam filter you use.

      For Spam-Assassin with the Bayes plugin, you can apply different scores to BAYES_00 up to the BAYES_99 tags and sort by score. Note that you have to add the X-Spam header to gnus-extra-headers to score them efficiently.

      Otherwise, it’s not difficult to define your own sort functions dependent on some header value. The following two examples should give you an idea how achieve that:

      Bogofilter:

      http://www.emacswiki.org/cgi-bin/wiki/GnusBogofilter

      CRM114:

      http://www.emacswiki.org/cgi-bin/wiki/GnusMailreaver

    • http://lifelogs.com Ted Zlatanov

      I maintain spam.el, the Gnus anti-spam library, and welcome any questions or suggestions in the Gnus mailing list, newsgroup, or here.

      You can sort message by their spam score, which is calculated specifically for every backend that supports it. For instance, I use


      (defun tzz-gnus-override-spam (group)
      (setq
      gnus-article-sort-functions
      '(spam-article-sort-by-spam-status)))

      for my spam groups, but not for regular groups (in regular groups, I sort by article number). This works for Bogofilter, CRM114, and others.

  • http://sachachua.com Sacha Chua

    That sounds like a job for scoring. =) More on that later!

  • deb

    Hello,

    I use gnus and would like to install a filtering spam method as those described here. But it is not clear enough how to proceed, what to install and
    how to configure,

    would you have a very basic configuration to start with ?
    with bogofilter, something that would learn from mails manually
    marked as spam ? and put it into a local spam folder

    I use gnus in conjunction with dovecot and offlineimap,
    I maintain an imap server on my laptop, which is synchronized with
    that of my ISP, then gnus asks this local server.

    Thanks in advance for your help

    Best regards

  • http://sachachua.com Sacha Chua

    Unfortunately, I haven’t gotten around to re-setting up Gnus on my Windows computer. Would you like to check the Emacs Wiki or ask in the Gnus user newsgroup?