# Configuration for CRM114 plugin # # CRM114 INSTALLATION & CONFIGURATION: # To use this plugin you have to set up CRM114 so that you have these files: # mailreaver.crm, mailfilter.cf, rewrites.mfp, priolist.mfp, and .CSS files # (see http://crm114.sourceforge.net/CRM114_Mailfilter_HOWTO.txt for details). # # The most important steps are: # mkdir ~/.crm114 # cp mailfilter.cf rewrites.mfp *.crm ~/.crm114 # cd ~/.crm114 # cssutil -b -r spam.css # cssutil -b -r nonspam.css # touch priolist.mfp # $EDITOR mailfilter.cf # $EDITOR rewrites.mfp # # In mailfilter.cf check the option ":add_headers: /yes/"! # (and do not bother to change the flag_subject_string options -- # this plugin ignores them anyway) # # Plugin CONFIGURATION: # To use the plugin you probably have to set the crm114_command. # All other settings should have working default values, # which are chosen to be cautionary and nonintrusive. # The commented examples for each settings are my own tested configuration. # # Further Notes on Licence and Implementation are in crm114.pm and its POD # ############################################################################# # these two lines are necessary to activate the plugin: # for older versions <0.7 # loadplugin crm114 crm114.pm # for newer versions >=0.7 loadplugin Mail::SpamAssassin::Plugin::CRM114 crm114.pm full CRM114_CHECK eval:check_crm() # this high priority is not necessary. but running late allows us # to compare the CRM score and the result of all previous SA tests # # 899 is chosen as an optimization because FuzzyOCR runs at 900 # thus if CRM already yields a high SA score, # then FuzzyOCR will decide to skip its tests priority CRM114_CHECK 899 # commandline to execute CRM114 # default: crm -u ~/.crm114 mailreaver.crm #crm114_command /usr/local/bin/crm -u /var/amavis/.crm114 mailreaver.crm # let SA add header lines to processed mails #add_header all CRM114-Version _CRM114VERSION_ #add_header all CRM114-CacheID _CRM114CACHEID_ add_header all CRM114-Status _CRM114STATUS_ ( _CRM114SCORE_ ) # ignore existing X-Spam or X-Virus headers # if SpamAssassin is called by Amavis then use the same value as Amavis does. # that way a SA-check from Amavis and on from the command line both see the same # Headers # default: 0 #crm114_remove_existing_spam_headers 1 #crm114_remove_existing_virus_headers 1 # dynamic score # values: 0 - returns subtest results # 1 - returns a dynamic CRM score (default) #crm114_dynscore 1 # dynamic score normalization factor # CRM score have much higher absolute values and different signs than SA scores # (usual ham-scores are between 15 and 40, scores from -10 to 10 are undecided, # previously seen spam easily gets -200). # With dynamic scoring the SA score is calculated by: * crm114_dynscore_factor # # Notes: - this has to be a negative number! # - the absolute value should be quite low (certainly <.3, probably <=.2), # otherwise the returned score would override all other tests. # default: calculate factor so that CRM-score -25 yields the SA required spam threshold #crm114_dynscore_factor -0.05 # static scores # without dynamic scores these scores are used # default values are respectively -3, 3, -0.5, 0.5, and 0 for good, spam, # probably good, probably spam, and unsure. good/spam depend on crm114's # classification the probably good/spam values are used # for |crm114-score| >= 5 (i.e. a crm114-classification of unsure) # leaving |crm114-score| < 5 for 'really' unsure #crm114_staticscore_good -3.0 #crm114_staticscore_prob_good -0.5 #crm114_staticscore_unsure 0.0 #crm114_staticscore_prob_spam 0.5 #crm114_staticscore_spam 3.0 # custom crm114 thresholds; these settings override variables :good_threshold: # and :spam_threshold: in mailfilter.cf, and are also used to determine # additional classes prob_good/prob_spam when crm114_dynscore is false. # default values are +10 for good threshold and -10 for spam threshold #crm114_good_threshold 10 #crm114_spam_threshold -10 # should CRM114 be trained by SA? # If enabled, then a call to Mail::SpamAssassin->learn() or # "spamassassin --report/--revoke" also calls the CRM114 plugin. # Since CRM114 uses a "Train On Error" strategy the plugin will check the # reported mail and only learn it if it is not classified correctly. # default: 0 #crm114_learn 1 # should CRM114 be trained by SA-autolearn? # If enabled, then SA's autolearn also calls the CRM114 plugin. # # This is different from :automatic_training: in CRM114's mailfilter.cf # because SA's score is influenced by several different factors while # CRM114 has to rely on its own classification. # But anyway: Only activate this if you know what you're doing! # (in other words: if non-learning SA rules (without AWL and Bayes) are # already well tuned and are known to provide good results) # default: 0 #crm114_autolearn 1 # should we preserve the message and its CRM114-CacheID for training # or discard it? # # to let CRM114 store messages into reaver_cache and use the cache for # manual learning enable it in mailfilter.cf, set this option, and # include the CacheID into all Mails with # "add_header all CRM114-CacheID _CRM114CACHEID_" # -- otherwise disable this option to strip CacheIDs before training # default: 0 #crm114_use_cacheid 1 # if both crm114_lookup_cacheid and crm114_use_cacheid are true # and CRM114-CacheID is not found in the message, do a lookup in the # reaver_cache/texts directory; crm114_cache_dir also needs to be set # default: 0 #crm114_lookup_cacheid 1 # a reaver_cache/texts directory used to lookup cacheid # if crm114_lookup_cacheid is set # default: ~/.crm114/reaver_cache #crm114_cache_dir path # should we skip CRM114 if other tests indicate certain spam/ham? # # disable CRM114 if a message already has a score (from other tests) # less than crm114_autodisable_negative_score or # more than crm114_autodisable_score. # # default: -999/999 # crm114_autodisable_negative_score -999 # crm114_autodisable_score 999