Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Added comments on how to tune the plugin

...

Please note that, while this plugin realises some sort of 'fuzzy checksum', the fuzzyness is realised by reducing the body text of an email to its (characteristic) minumum. The MD5 hashing algorithm used is not fuzzy. Accordingly there are no 'confidence levels' or things of that sort. It's hit or no hit.

There are likely to be FPs, especially with relatively short emails. You can somewhat tune the plugin by editing the RE with the code - see comments for that.

For those more deeply interested: This plugin is based on parts of the procmail-based project 'NiX Spam', developed by Bert Ungerer.(un@ix.de) If you can read German, read up at http://www.heise.de/ix/nixspam/. The procmail code producing the hashes only can be found here: ftp://ftp.ix.de/pub/ix/ix_listings/2004/05/checksums

...

No Format
=head1 NAME

Mail::SpamAssassin::Plugin::iXhash - compute fuzzy checksums from mail bodies and compare to known spam ones via DNS

=head1 SYNOPSIS
  loadplugin    Mail::SpamAssassin::Plugin::iXhash /path/to/iXhash.pm
  ixhash_timeout	10
  body          IXHASH eval:ixhashtest('ix.dnsbl.manitu.net')
  describe      IXHASH This mail has been classified as spam @ iX Magazine, Germany
  tflags        IXHASH net
  score         IXHASH 1.5

=head1 DESCRIPTION

iXhash.pm is a plugin for SpamAssassin 3.1.0 and up. It takes the body of a mail, strips parts from it and then computes a hash value from the rest.
These values will then be looked up via DNS. 
This plugin is based on parts of the procmail-based project 'NiX Spam', developed by Bert Ungerer.(un@ix.de)
For more information see http://www.heise.de/ix/nixspam/. The procmail code producing the hashes only can be found here:
ftp://ftp.ix.de/pub/ix/ix_listings/2004/05/checksums

Parts of the code were submitted via heise forum by 'kungfuhasi'
See http://www.heise.de/ix/foren/go.shtml?read=1&msg_id=7246759&forum_id=48292.

Martin Blapp (mb@imp.ch) found and solved a problem occuring on Perl 5.8.7. Thanks a lot!

Further improvements (DNS timeouts) by Dallas Engelken (dallase@uribl.com) - see comments for details. 

=cut

package Mail::SpamAssassin::Plugin::iXhash;
use strict;
use Mail::SpamAssassin;
use Mail::SpamAssassin::Plugin;
use Mail::SpamAssassin::Util;
use Digest::MD5 qw(md5 md5_hex md5_base64);
use Net::DNS;
use Net::DNS::Resolver;
# Locale - this was on Bert's wishlist
use POSIX qw(locale_h);
setlocale(LC_CTYPE, "de_DE.ISO8859-1");
# LC_CTYPE now "Deutsch, Deutschland, codeset ISO 8859-1"
# Maybe not appropriate for spam that is neither German nor English

use vars qw(@ISA);
@ISA = qw(Mail::SpamAssassin::Plugin);
sub dbg { Mail::SpamAssassin::dbg (@_); }
sub new {
        my ($class, $mailsa, $server) = @_;
        $class = ref($class) || $class;
        my $self = $class->SUPER::new($mailsa);
        bless ($self, $class);
        $self->set_config($mailsa->{conf});
        $self->register_eval_rule ("ixhashtest");
        return $self;
}


sub set_config {
	my ($self, $conf) = @_;
	my @cmds = ();
	# implements ixhash_timeout config option - by dallase@uribl.com
	push(@cmds, {
		setting => 'ixhash_timeout',
		default => 5,
		type => $Mail::SpamAssassin::Conf::CONF_TYPE_NUMERIC,
        });
	$conf->{parser}->register_commands(\@cmds);
}


sub ixhashtest {
	my ($self, $permsgstatus,$muell,$dnsserver) = @_;
        dbg("IXHASH: IxHash querying Server $dnsserver");
        my ($digest,$answer,$ixdigest,$body) = "";
        my @body = $permsgstatus->{msg}->get_body();
        my $resolver = Net::DNS::Resolver->new;
        my $body_copy = "";
        foreach (@body) {
                $body .= join "", @$_;
        }
        my $rr;
        my $hits = 0;
        # alarm the dns query - dallase@uribl.com
        # --------------------------------------------------------------------------
        # here we implement proper alarms, ala Pyzor, Razor2 plugins. 
        # keep the alarm as $oldalarm, so we dont loose the timeout-child alarm
        # see http://issues.apache.org/SpamAssassin/show_bug.cgi?id=3828#c123
        my $oldalarm = 0;
        my $timeout = $permsgstatus->{main}->{conf}->{'ixhash_timeout'} || 5;
        eval {
		Mail::SpamAssassin::Util::trap_sigalrm_fully(sub { die "ixhash timeout reached"; });
	$oldalarm = alarm($timeout);
	
        #-----------------------------------------------------------------------
        # Creation of hash # 1 if following conditions are met:
        # - mail contains at least 16 spaces or tabs 
	# - mail consists of at least 2 lines
      # NB:  Edit this if you want to minimize FPs at the cost of not checking shorter mails.
      # FP ratio will be the higher the shorter the mails are 
        if (($body =~ /([\s\t].+?){16,}/ ) && ($body =~ /.*$.*/)){
                # Copy $body into $body_copy - thanks to J. Stehle for pointing this out
                $body_copy = $body;
                # All space class chars just one time
		# Do this in two steps to avoid Perl segfaults
                # if there are more than 2.600 identical chars to be replaced
		# Step One
                $body_copy =~ s/([[:space:]]{100})(?:\1+)/$1/g;
		# Step Two
                $body_copy =~ s/([[:space:]])(?:\1+)/$1/g;
                # remove graph class chars and some specials
                $body_copy =~ s/[[:graph:]]+//go;
                # Create actual digest
                $digest = md5_hex($body_copy);
                dbg ("IXHASH: Computed hash-value $digest via method 1");
                dbg ("IXHASH: Now checking $digest.$dnsserver");
                # Now check via DNS query
                $answer = $resolver->search($digest.'.'.$dnsserver, "A", "IN");
                if ($answer) {
                        foreach $rr ($answer->answer) {
                                next unless $rr->type eq "A";
                                dbg ("IXHASH: Received reply from $dnsserver:". $rr->address);
                                $hits = 1 if $rr->address;
                        }
                }
        }
	# End of computing hash #1 
	#---------------------------------------------------------------------
        
	
	#---------------------------------------------------------------------
        # Creation of hash # 1 if mail contains at least 3 of the following characters:
        # '[<>()|@*'!?,]' or the combination of ':/'
        # (To match something like "Already seen?  http:/host.domain.tld/")
        # edit this if you want to minimize FPs (i.e. make sure that short emails are not checked)
        # 
        if ($body =~ /((([<>\(\)\|@\*'!?,])|(:\/)).*?){3,}/m ) {
                $body_copy = $body;
		# remove redundant stuff
                $body_copy =~ s/[[:cntrl:][:alnum:]%&#;=]+//g;
		# replace '_' with '.'
                $body_copy =~ tr/_/./;
                # replace duplicate chars. This too suffers from a bug in perl
		# so we do it in two steps
		# Step One
                $body_copy =~ s/([[:print:]]{100})(?:\1+)/$1/g;
                # Step Two
                $body_copy =~ s/([[:print:]])(?:\1+)/$1/g;
		# Computing hash...
                $digest = md5_hex($body_copy);
                dbg ("IXHASH: Computed hash-value $digest via method 2");
                dbg ("IXHASH: Now checking $digest.$dnsserver");
                # Now check via DNS query
                $answer = $resolver->search($digest.'.'.$dnsserver, "A", "IN");
                if ($answer) {
                        foreach $rr ($answer->answer) {
                                next unless $rr->type eq "A";
                                dbg ("IXHASH: Received reply from $dnsserver:". $rr->address);
                                $hits = 1 if $rr->address;
                        }
                }
        }

	# End of computing hash #2 
	#-----------------------------------------------------------------------
        
	
	#-----------------------------------------------------------------------
	# Compute hash # 3 if 
        # - there are at least 8 non-empty characters in the body
	# - neither hash #1 nor hash #2 have been computed
        if (($body =~ /[\S]{8,}/) and (length($digest) < 32)) {
                $body_copy = $body;
                $body_copy =~ s/[[:cntrl:][:space:]=]+//g;
                # replace duplicate chars. This too suffers from a bug in perl
		# so we do it in two steps
		# Step One
                $body_copy =~ s/([[:print:]]{100})(?:\1+)/$1/g;
                # Step Two
                $body_copy =~ s/([[:graph:]])(?:\1+)/$1/g;
		# Computing actual hash
                $digest = md5_hex($body_copy);
                dbg ("IXHASH: Computed hash-value $digest via method 3");
                dbg ("IXHASH: Now checking $digest.$dnsserver");
                # Check via DNS
                $answer = $resolver->search($digest.'.'.$dnsserver, "A", "IN");
                if ($answer) {
                        foreach $rr ($answer->answer) {
                                next unless $rr->type eq "A";
                                dbg ("IXHASH: Received reply from $dnsserver:". $rr->address);
                                $hits = 1 if $rr->address;
                        }
                }
        }
        if (defined $oldalarm) {
		alarm $oldalarm;  $oldalarm = undef;
        }
}; # End of sub ixhashtest

# Error handling - parto of Dallas' code
if ($@ =~ m/timeout/) {
	dbg("IXHASH: $timeout second timeout exceeded while checking $digest.$dnsserver");
}

return $hits;

}
1;

...