Print Page | Close Window

Import messages

Printed From: LogSat Software
Category: Spam Filter ISP
Forum Name: Spam Filter ISP Support
Forum Description: General support for Spam Filter ISP
URL: https://www.logsat.com/spamfilter/forums/forum_posts.asp?TID=5028
Printed Date: 22 December 2024 at 3:39pm


Topic: Import messages
Posted By: Derk
Subject: Import messages
Date Posted: 26 January 2005 at 4:24pm

is it, or could it be, possible to have the spam filter application read a text file containing e-mail messages to be analyzed and included in the Bayesian filter database? after they are imported the application could erase the file.

I would like to be able to forward messages that get by the filter to a mailto:spam@mydomain - spam@mydomain catch all account.  Then I can set up a scheduled task to retrieve the messages and write them to a text file to be imported by the application.




Replies:
Posted By: LogSat
Date Posted: 26 January 2005 at 8:34pm
Derk,

That is currently not possible. SpamFilter does scan a folder (\SpamFilter\corpus\queue\*.token) for temporary text files  with a .token extension. However the token files must only contain the email's "tokens", not the full email text. Tokens are the keywords that SpamFilter extracts from the email itself. The 1st line of the text file containt the words ".good" or ".spam" or ".falsepositive" to catalog the tokens as belonging to a good or spam email.
In theory you could write an application to create tokens out of an email, but we're not going to be able to provide any details on how and using which rules we use to create tokens from an email. We don't want any spammer to find out about the innerworkings of our software...


-------------
Roberto Franceschetti

http://www.logsat.com" rel="nofollow - LogSat Software

http://www.logsat.com/sfi-spam-filter.asp" rel="nofollow - Spam Filter ISP


Posted By: Web123
Date Posted: 27 January 2005 at 1:30am
I also get quite a lot false positives trough my system, and would
love to "import" them somehow to Spamfilter?
 
regards,
Kim


Posted By: Marrab
Date Posted: 27 January 2005 at 1:38am

Thanks. Now i can although somehow control bayesian filter. One more question: does any quarantined mail be considered bayesian filter as spam?



Posted By: Marrab
Date Posted: 27 January 2005 at 1:52am
It seems i understand why my bayesian filter working so strange. Token files contains only mail headers. Interestingly why?


Posted By: Derk
Date Posted: 27 January 2005 at 4:20pm

Originally posted by LogSat LogSat wrote:

Derk,

That is currently not possible. SpamFilter does scan a folder (\SpamFilter\corpus\queue\*.token) for temporary text files  with a .token extension. However the token files must only contain the email's "tokens", not the full email text. Tokens are the keywords that SpamFilter extracts from the email itself. The 1st line of the text file containt the words ".good" or ".spam" or ".falsepositive" to catalog the tokens as belonging to a good or spam email.
In theory you could write an application to create tokens out of an email, but we're not going to be able to provide any details on how and using which rules we use to create tokens from an email. We don't want any spammer to find out about the innerworkings of our software...

what about if I set up a subdomain that I can forward messages to, like mailto:spam@junk.mydomain.com - spam@junk.mydomain.com ?  Can I configure the application to consider every single message spam to be added to the Bayesian database?



Posted By: LogSat
Date Posted: 27 January 2005 at 11:16pm
We recommend strongly against "interfering" with the Bayesian filter by "forcing" emails or worse spam thru it. The bayesian filter needs to see the true email received by the mail server to function properly. If junk mail is fed to it, that junk is not representitative of the true emails received by SpamFilter, and the bayesian filter will loose its accuracy.

To answer Marrab, yes, all email that is quarantined is considered to be spam by the statistical filter.


-------------
Roberto Franceschetti

http://www.logsat.com" rel="nofollow - LogSat Software

http://www.logsat.com/sfi-spam-filter.asp" rel="nofollow - Spam Filter ISP


Posted By: Web123
Date Posted: 28 January 2005 at 1:20am

How about at least Blacklist all emails sent to mailto:spam@junk.mydomain.com - spam@junk.mydomain.com ? Could it be done?



Posted By: LogSat
Date Posted: 28 January 2005 at 5:06pm
Yes, it can:

  • file:///c:/My%20Documents/Delphi%20Projects/SpamFilter/readme.html#BlacklistedTOEmails - - The recipient's email address is checked against your local list of blacklisted email addresses. If present, it is rejected.



-------------
Roberto Franceschetti

http://www.logsat.com" rel="nofollow - LogSat Software

http://www.logsat.com/sfi-spam-filter.asp" rel="nofollow - Spam Filter ISP


Posted By: Web123
Date Posted: 29 January 2005 at 2:54am
sorry!,
meant that if I get a spam-message that is not caught by the filter,
I could then forward it to a address( mailto:spam@junk.mydomain.com - spam@junk.mydomain.com )
and it would parse the message and blacklist all the non local addressees.
 
Kim


Posted By: Guests
Date Posted: 30 January 2005 at 8:28am
In theory you could, but you would have to forward the original, unmodified message. This means that the email headers MUST be exactly as the original ones, and the same applies to the email body. Sipmly taking the email and forwarding it using an email client will not work accurately, as email clients, especially Outlook, completely modiffy the original email's format.


Posted By: Guests
Date Posted: 01 March 2005 at 9:42am
We host many domains and some clients as that we bypass the filter for their domain.  When we pass their mail though the filter does this impact our bayesian filter or are these emails not considered in the filering rules?

It seems the longer we run our system the less effective the bayesian filter becomes we are considering dumping the database and starting over but we have process 500000 plus emails and hate to lose all those statistics?

Any help would be really apprecitated.



Print Page | Close Window