My Spam Filter Program 
Author Message
 My Spam Filter Program

To All,

Spam is a major problem, and it will probably get worse. After looking at
programs such as SpamAssassin and Bayesian filtering, I decided to try
simpler methods of analyzing the header and first few lines of the body.

The program is remarkably accurate in sorting good messages from spam,
and runs between 300 to 500 messages per second, depending on the message
size. It works in dos and uses the plain ascii format of the Netscape
Inbox (versions 2.02 to 4.0). It should work with any client that saves
messages in one file and preserves the headers.

Others may be interested in adapting the program to their own needs, so I
have uploaded it to

  http://www.*-*-*.com/

Mike Monett



Tue, 04 Oct 2005 15:29:44 GMT  
 My Spam Filter Program
Oops - a bug.

In the file "spammain.pas", there is a new routine that locates the start
of a line and copies it to a string variable. The section is shown below:

procedure NextLine; assembler;


  mov  [offset Rsv + bx], al            {copy char to Rsv}

  add  bx, 1                            {point to next char in Rsv}

BX is a word, so the routine will not detect the end of the string and
happily write past it. The program hangs on long lines.

Please change the line to

  add  bl, 1                            {point to next char in Rsv}

This works better.

I just sorted 333 spam messages in 1 second in a file 1.3 megabytes long.
Two of the messages should have been declared spam, but they were done so
well they looked like ordinary email.

The accuracy is only 99.39%, but I'm quite pleased with the performance.
So far, it has not declared a single valid email as spam.

I actually look forward to receiving spam for the pure pleasure of
killing it!

Best Regards,

Mike Monett



Tue, 04 Oct 2005 17:03:09 GMT  
 My Spam Filter Program

Quote:

> To All,

> Spam is a major problem, and it will probably get worse. After looking at
> programs such as SpamAssassin and Bayesian filtering, I decided to try
> simpler methods of analyzing the header and first few lines of the body.

> The program is remarkably accurate in sorting good messages from spam,
> and runs between 300 to 500 messages per second, depending on the message
> size. It works in dos and uses the plain ascii format of the Netscape
> Inbox (versions 2.02 to 4.0). It should work with any client that saves
> messages in one file and preserves the headers.

> Others may be interested in adapting the program to their own needs, so I
> have uploaded it to

>   http://www3.sympatico.ca/add.automation/misc/spa2e921.zip

> Mike Monett

I have uploaded the most recent version. It now decodes over 2,200
messages/sec on small files, and >300 messages/sec on large files.

There are no false positives or false negatives.

Long Live Pascal!

Mike Monett



Fri, 07 Oct 2005 05:16:13 GMT  
 
 [ 5 post ] 

 Relevant Pages 

1. Anti Spam and Spam Proofing

2. Free SPAM Program

3. Free SPAM Program

4. Free SPAM Program

5. Spam Detection By Analyzing Headers

6. Steve Spam-bot Thank you

7. Away with SPAM - solution!

8. Away with SPAM - solution!

9. Away with SPAM - solution!

10. It's SPAM time

11. Away with SPAM - solution!

12. How to filter a Lookup directly without filtering the Table

 

 
Powered by phpBB® Forum Software