iXhash.pm is a plugin for SpamAssassin 3.1.0 and up (including 3.2, as far as I can tell). A version compatible with SpamAssassin 3.0 is provided as well.
Basically the plugin provides a network-based test, just as razor2, pyzor and DCC do. Working solely on the body of an email, it removes parts of it and computes a hash value from the rest. These values will then be looked up via DNS using the domains given in the config file(s). You need Net::DNS and Digest::MD5 installed.
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
Instructions
The top file is ixhash.cf - an example config file
The middle file 'ixhash.pm' is the current version, use this with SA 3.1.x
The bottom file is 'ixhash_v300.pm', use this if you run SA 3.0.x.
- copy paste either 'ixhash.pm' or 'ixhash_v300.pm' into a file called ixhash.pm, depending on the SA version you plan to use the plugin with. 2. copy-paste the text contained in 'ixhash.cf' into a file called "ixhash.cf" 3. place ixhash.pm somewhere where SpamAssassin can access it - like /etc/mail/spamassassin. 4. edit the first line of ixhash.cf (loadplugin) to reflect where ixhash.pm is. 5. if you have to use the ixhash_v300.pm version, delete/comment the 'ixhash_timeout' line. This feature is not available with your version 6. copy ixhash.cf into /etc/mail/spamassasssin 7. run 'spamassassin -D < /some/mail/message', where '/some/mail/message' is a file containing an email. The copious output (on STDERR) should contain some occurences of 'IXHASH', showing you the plugin is being used. 8. restart spamd/amavisd/whateverDaemonUsesSpamAssassin
ixhash.cf
loadplugin Mail::SpamAssassin::Plugin::iXhash /path/to/iXhash.pm # This makes DNS queries time out after 10 seconds (2x default) ixhash_timeout 10 # This list uses iX Magazine's spam as datasource. 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 # This list comes in @ spamtraps run by LogIn & Solutions AG, Germany # Manually verified stuff body LOGINHASH1 eval:ixhashtest('nospam.login-solutions.de') describe LOGINHASH1 mail has been classified as spam @ LogIn&Solutions AG, Germany tflags LOGINHASH1 net score LOGINHASH1 1.5 # This list contains hashes from Mails classified as spam at a larger company based in Germany # Lots of stuff, but automatically categorized and contributed body LOGINHASH2 eval:ixhashtest('nospam.login-solutions.ag') describe LOGINHASH2 mail has been classified as spam @ unknown company, Germany tflags LOGINHASH2 net score LOGINHASH2 1.5
ixhash.pm
=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;
ixhash_v300.pm
=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 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 3.0.x. 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! If you use SpamAssassin 3.1.0 and higher, please use the code above (in ixhash.pm), not this one. Dallas Engelken has introduced some code that offers the capability to set an timeout limit for DNS queries but requires 3.1.x. If you can you should use that version. This code here is provided only for compatibility with SA 3.0 (as found on current Debian System). Naturally you cannot set a ixhash_timeout configuration value when using this version. =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->register_eval_rule ("ixhashtest"); return $self; } sub ixhashtest { my ($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 $hits = 0; #----------------------------------------------------------------------- # Creation of hash # 1 if following conditions are met: # - mail contains at least 16 spaces or tabs # - mail consists of at least 2 lines if (($body =~ /([\s\t].+?){16,}/ ) && ($body =~ /.*$.*/)){ $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 $rr->address; } } } # End of computing hash #1 #--------------------------------------------------------------------- #--------------------------------------------------------------------- # Creation of hash # 1 if mail contains at least 3 of the following characters: # '[<>()|@*'!?,]' or the combination of ':/' # (To match something like "Already seen? http:/host.domain.tld/") if ($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; } } }; # End of sub ixhashtest return $hits; } 1;