SpamAssassin - How Do I Train?

[rmmoore80]

My 7.5 installation of Plesk for Windows seems to be doing good but I can't figure out how to train SpamAssassin. It does not work. Is it missing or is this module not working in 7.5?

 

[superm0nk]

I'd also like to know this.

 

[rmmoore80]

There HAS to be a way, has anyone gotten SpamAssassin to work with a default installation? I really need to get this working.

 

[webbit]

no spam assasin training icons

I am also missing the Spam Assasin Training icons.

I can change the white and black list and the hit level but no icon for training SA.

Does anyone have this feature working in 7.5.2 for windows?

Thanks

 

[rmmoore80]

Does yours even work at all? Mine doesn't...

I found out how to train...but I still can't get it to work. I need more Spam and Ham files. I need 200 spam mail and 200 ham (good) mail. You have to use sa-learn --spam c:\SpamLocation from your servers command prompt where the location is a directory you are storing your sample spam in. If someone can find me 200 pieces of spam AND 200 pieces of good mail in MailEnable's *.mai format then I can make a writeup on how to do this if I can get it to work. I have 130 spam trained and 30 ham trained but it still does not work.

 

[rmmoore80]

I found a database of spam, about 3000 Spam and 3000 NonSpam. I'll test it out and if I can get mine to work I will create a write up and post it here. I'll also send it to SWsoft so they can include it with the documentation so hopefully people can do this easier in the future.

 

[renhack]

Where did you find the spam/ham? I've been looking for some sample spam for a while.

 

[rmmoore80]

I went to google.com and did a search for:

Directory of "Sample Spam" and got this url:

http://ftp.novell.com/pub/forge/nw-assp/Sample Spam Database/

It's a tar zip file that contains two directories, one is Spam the other is NonSpam. About 2800 of each. Be sure you rename the file from *.tgz.gz to *.tar.gz or Winzip won't open it.

Download that on your server. Then use the SpamAssassin help files to figure out how to do training. Ie: From command prompt in the SpamAssassin directory type sa-learn --spam d:\Spam for Spam or sa-learn --nonspam d:\NonSpam for NonSpam. Replace location with what you need.

In my case, I stored my Spam in d:\spam and nonspam in d:\NonSpam.

 

[renhack]

Thanks for the followup.

I allready have the spam training knowledge. Worked as an email admin for an isp previous to the hosting industry. I just have slim pickins for HAM examples.

This helps alot.

 

[rmmoore80]

You're welcome. PLEASE let me know if yours starts flagging Spam or not. I have yet to get any spam since running the training, so I haven't had the opportunity to see.

 

[renhack]

I just finished training. Im using the spambayes plugin for outlook so I'll have to turn that off for a few days. I'll keep you posted.

 

[rmmoore80]

Thank you. I will do the same and keep everyone here posted.

 

[rmmoore80]

Anyone get this to work yet? I've trained 3k spam and 3k nonspam and it still isn't flagging anything, correctly or not.

 

[SecondPhase]

individual user training

Hi, i'm trying to so some traninig on a new SA installed server. sa-learn from the command line looks like it will work with a system wide setup since it looks for files in ~/.spamassassin but with the Plex/Qmail multiuser email setup, each user has Bayes files and the --username option does not seem to work with qmail accounts..

One idea might be to run sa-learn on a system users to create bayes-.. files then copy them to each user location. Or can system wide bayes- files from another server by copies on?? anybody know if this is a good idea or will cause problems?

Thanks!

FYI: i'm not on a windows machine but Plesk/Fedora 2 didn't realize i had latched a thread in the windoze forum.

 

[atinoco]

Hello guys, I'm kinda new to spamassassin, but im very interested in fine tuning it for my spam fighting needs.

I've setup SA via plesk to mark spam with a 5 score, also installed some SARE SA Rules and they are working fine. But I get false positives every once in a while

I've been reading about training for quite a while but i havent tried to do it yet.

I understand that I require samples of spam and ham messages to do this, my question its this, how do i extract these spam samples from Outlook 2003, anybody have any experience with this?

Thanks in Advance

 

[atinoco]

Some more info:

I also don't have spamassassin enabled serverwide, I enabled it only to "personal settings".

is it possible to specify to which email box the training is done? my plan its to collect spam & ham samples from my personal email and then just train SA for my personal needs (just for my email box). I dont want my bayessian trainign to affect the other perople hosted in my server

I run Plesk 7.5.6 on Windows 2003 Server

Thanks in Advance

 

[jovana]

Originally posted by atinoco

I've setup SA via plesk to mark spam with a 5 score, also installed some SARE SA Rules and they are working fine.

Hi,

What did you do to make this rules working? In which directory did copy this rules files?

thx