How to tell if SpamAssassin Bayasian filtering is working

Wed Sep 27 12:12:49 IST 2006

Hello Philip,

OK, so I now see that there are only about 127 hams.

I also have autolearn turned on.  Why isn't it picking up on the spams that it currently detects and adding them to the database?

Thanks in advance

JBB

Tuesday, September 26, 2006, 12:50:30 PM, you wrote:

PZ> Run the --lint test to see how many ham and spam messages it has.
PZ> Spamassassin needs to have at least 200 spam and 200 ham in bayes before it
PZ> considers the bayes to be trained.  It ignores the bayes results until then.

PZ> Philip

PZ> -----Original Message-----
PZ> From: mailscanner-bounces at lists.mailscanner.info
PZ> [mailto:mailscanner-bounces at lists.mailscanner.info] On Behalf Of Jonathan B.
PZ> Bayer
PZ> Sent: Tuesday, September 26, 2006 11:15 AM
PZ> To: MailScanner discussion
PZ> Subject: How to tell if SpamAssassin Bayasian filtering is working

PZ> Hello MailScanner,

PZ> I've enabled Bayasian filtering with SpamAssassin.  I've been feeding missed
PZ> spam to Spamassassin by hand, and I get reports that it is adding tokens
PZ> from the scanned files.  

PZ> However, it doesn't seem to have affected the results I'm getting.  I've
PZ> looked both at spam and non-spam, and don't see any reference in the headers
PZ> to any sort of Bayasian scanning.  

PZ> So I'm wondering if it is really doing the Bayasian scanning or not.

PZ> I've reproduced all the relevent information below (I think).  Any ideas or
PZ> suggestions are welcome.

PZ> Thanks in advance.

PZ> JBB

PZ> I have the following relevant lines in my MailScanner.conf file:

PZ> Spam Checks = yes
PZ> Use SpamAssassin = yes
PZ> Wait During Bayes Rebuild = no
PZ> SpamAssassin Site Rules Dir = /etc/mail/spamassassin
PZ> SpamAssassin Site Rules Dir = /etc/mail/spamassassin
PZ> SpamAssassin Local Rules Dir = 
PZ> SpamAssassin Local State Dir = # /var/lib
PZ> SpamAssassin Default Rules Dir =
PZ> Debug SpamAssassin = no

PZ> There is a soft link in place:
PZ> /etc/mail/spamassassin/mailscanner.cf -> 
PZ>          /etc/MailScanner/spam.assassin.prefs.conf

PZ> The file spam.assassin.prefs.conf contains the following:

PZ> dns_available yes
PZ> use_bayes 1
PZ> bayes_path /var/MailScanner/bayes/bayes
PZ>  score BAYES_00 -15.0
PZ>  score BAYES_05 -5.0
PZ>  score BAYES_95 5.0
PZ>  score BAYES_99 15.0
PZ>  bayes_auto_learn 1
PZ> bayes_ignore_header bayerfamily.net-MailScanner
PZ> bayes_ignore_header bayerfamily.net-MailScanner-SpamCheck
PZ> bayes_ignore_header bayerfamily.net-MailScanner-SpamScore
PZ> bayes_ignore_header bayerfamily.net-MailScanner-Information
PZ> bayes_auto_expire 0
PZ> lock_method flock
PZ> use_auto_whitelist 0
PZ> use_dcc         0
PZ> use_pyzor       0
PZ> envelope_sender_header X-MailScanner-From
PZ> header   FRIEND_GREETINGS       Subject =~ /you have an E-Card from/i
PZ> describe FRIEND_GREETINGS       Nasty E-card from FriendGreetings.com
PZ> score    FRIEND_GREETINGS       100.0
PZ> header   FRIEND_GREETINGS2      Subject =~ /you have a greeting card from/i
PZ> describe FRIEND_GREETINGS2      Nasty E-card from FriendGreetings.com
PZ> score    FRIEND_GREETINGS2      100.0

PZ> And /var/MailScanner/bayes contains:
PZ> -rw-------  1 root root    230 Sep 25 04:03 bayes.mutex
PZ> -rw-------  1 root root  49152 Sep 25 04:03 bayes_seen
PZ> -rw-------  1 root root 360448 Sep 25 04:03 bayes_toks

PZ> JBB
PZ> ---
PZ> Jonathan B. Bayer

PZ> mailto:jbayer at bayerfamily.net

PZ> -- 
PZ> MailScanner mailing list
PZ> mailscanner at lists.mailscanner.info
PZ> http://lists.mailscanner.info/mailman/listinfo/mailscanner

PZ> Before posting, read http://wiki.mailscanner.info/posting

PZ> Support MailScanner development - buy the book off the website! 

PZ> -- 
PZ> This message has been scanned for viruses and
PZ> dangerous content by MailScanner, and is
PZ> believed to be clean.

JBB
---
Jonathan B. Bayer

mailto:jbayer at bayerfamily.net