ashesnglass Posted April 12, 2004 Posted April 12, 2004 Hi, I use Spam Assassin to filter tons of spam, but during the past week or two, it's been marking lots of legitimate mail as spam. In every one of these instances, it's been because of a high Bayesian spam score. How can I disable Bayesian filtering? I don't quite understand the instructions associated with "Score" in the Spam Assassin configuration tool. Thanks! Dave Quote
TCH-Don Posted April 12, 2004 Posted April 12, 2004 I had to lower the score to 6 to get my forms to pass through. Its worth a try, but some of the spam will get through then. But its a lot less than without spam assassin. Quote
Deverill Posted April 12, 2004 Posted April 12, 2004 An email service I use has a header that shows the score as *'s. I haven't looked here to see if it is standard or something they customized. I just set a filter where header is **** to my Spam4 folder and *** to my Spam3 folder. This way I know I can cruise through the Spam4's because there's little chance of something legit there and then take a second on the Spam3's because they're not as "spammy". I hope maybe this can work for you too. Quote
TCH-Bruce Posted April 12, 2004 Posted April 12, 2004 Don, I had the same problem with forms on our company site but I found a way around it while still leaving Spam Assassin set at level 5. You need to make some mods to your user_prefs file in the .spamassassin folder. I did it like this. header MSGID_FROM_MTA_SHORT Subject =~ /{Your Subject here}/ score MSGID_FROM_MTA_SHORT -5 This used a score for MSGID_FROM_MTA_SHORT of -5 for my subject matter. This drops the scoring on these messages by -5 and they get through leaving you set at a lower threshold for Spam Assassin. You could modify the value even more if you wanted and drop your threshold even more. It's trial and error I've found to find a happy medium. Quote
ashesnglass Posted April 13, 2004 Author Posted April 13, 2004 I changed the score to 6.5 but I'm still losing e-mails to the filter and tons are getting through. I want to disable the Bayesian thing completely. Can anybody tell me how to do this in CPanel? Thanks Quote
TCH-Bruce Posted April 13, 2004 Posted April 13, 2004 (edited) You will have to modify your user_prefs file in the .spamassassin folder. This folder is in your root folder. Edit the user_prefs file using cpanel's file manager. Add the following line. use_bayes 0 That will turn off Bayes filtering. Here is a link to the documentation for the user_prefs file. http://eu.spamassassin.org/doc/Mail_SpamAssassin_Conf.html Edited April 13, 2004 by TCH-Bruce Quote
hruska Posted April 20, 2004 Posted April 20, 2004 I turned of bayes via use_bayes=0 and it worked. However, my spam that got through went way up. I decided to turn it back on and am still, 5 days later, not getting a single BAYES rule hit in any email. The databse files in my .spamassassin directory were being updated, so at least that wasn't broken. But I figured 5 days would be long enough for things to be trained and start seeing BAYES rules applied. I since completelely disabled spamassasin, manually deleted the .spamassassin directory, and then re-enabled it. This was in hopes of reinitializing something that may have broken... Anyone know more about this? How long it might take? Is there any user level training one can do? I had used another bayes program in the past, Popfile, and it had you train it based on every message you thought was spam. I'm not sure how spamassassin handles this. Quote
hruska Posted April 20, 2004 Posted April 20, 2004 Hate to reply to my own post so soon, but I think I found some answers... It looks that by default spamassasin will not apply BAYES rules until it has seen 200 spam and 200 ham messages. I have gotten the spam number for sure, but I don't have that much ham in my database since I've cleared it. I do notice that it is flagging many messages per day as spam via the autolearn=spam headers. I will play with the 'bayes_min_ham_num' setting in my user prefs and see if things start to kick in. Quote
TCH-Bruce Posted April 20, 2004 Posted April 20, 2004 Yes, it has to relearn the rules before it will become effective. It needs a base to work from. Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.