Junk Mail Spam filter - Banned Subjects problem

[Bozwell2]

I have placed a list of words in the filter under banned subjects, however, I continue to get e-mail spam with those words. I am doing this in the Junk Mail section, under Word Lists, under Banned Subjects. Here is what my list looks like:

#" --- ENTER SUBJECTS BELOW THIS LINE ---
BuyBelowCost
below cost
rolex
dell
loans
price
smoking
crucial
wieght
megapixel
drugs
druugs
cicodin
low rates
free shipping
college degree
refinancing
viagra
debt
x-box
Am I doing something wrong? I have e-maild for support but received no reply.

Thanks,

Steve

[Michael]

The first thing to check is that you are running the non-Bayesian filters, "Tools | Junk Mail Filtering | General Settings".

[dribnus]

I'm not running that. I thought that was correct. If not, please let me know. What does that do?
Thanx.

[Bozwell2]

Under Junk Mail, General Settings, there is a checkmark next to Run non-Bayesian Filters. So that I believe is set properly.

Steve

[Michael]

I'm not running that. I thought that was correct. If not, please let me know. What does that do?
Thanx.

This adds Poco's built in junk mail filters (the ones Poco used prior to the introduction of built-in Bayesian filtering) in addition to the Bayesian filters.

[Michael]

Under Junk Mail, General Settings, there is a checkmark next to Run non-Bayesian Filters. So that I believe is set properly.

Steve:

Yes, this is the setting you want.

Do these messages end up in your IN box or some other mailbox? If it is not the IN box then it is likely the result of some other filter moving the messages. When a filter moves a message then no subsequent filter will be executed against that message.

Assuming this is not the case and that the message is ending up in your IN box then please try the following:

I have seen that sometimes the Bayesian filters are not executed on messages I receive. I've generally put this down to the fact that I have many levels of junk mail filtering in use. I thought that perhaps one of these prior levels might have been responsible. (I wasn't too worried as my junk mail filtering is proving very accurate).

Perhaps I am wrong in my thinking, maybe there are circumstances where the JMF filters are not being executed. Please try the following test. For one of the messages with a banned word in the subject select "Tools | Junk Mail Filtering" and then press the "Apply and Test" button. Do the filters report an item for "SUBJECT = %JUNK SUBJECTS%". If it does then check the examine the full message headers and check to see if the following header exists in the message:

Code: Select all

X-Poco-Score-Detail: +20 [SUBJECT=%JUNKSUBJECTS%] (Subject %junksubjects%)

In fact check for any "X-Poco-Score-Detail" header(s).

[Bozwell2]

I do not find that exact message, but here is what I find that has most of it:

X-Poco-Score-Detail: -100 [%BAYES%=P=50;T=90;BIAS=+20] (%bayes% P=50;T=90;Bias=+20)
X-Poco-Score-Detail: +3 [X-MAILER=] (X-Mailer )
X-Poco-Score-Detail: -4 [TO=%ADDRESSBOOKS%]OR[CC=%ADDRESSBOOKS%] (To %addressbooks%, or Cc %addressbooks%)
X-Poco-Score-Detail: +2 [FROM=%ADDRESSBOOKS%] (From %addressbooks%)
X-Poco-Score-Detail: +100 [SUBJECT=%JUNKSUBJECTS%] (Subject %junksubjects%)
X-Poco-Scored: +1


BTW, when I did the junk mail test, it said it would be considered junk mail under all circumstances. And yes, it is arriving in the In box. (thanks for your help with this - currently my junk mail filtering is only at 20% accuracy.)

Steve

Edit: Moved to Junk Mail Filtering by Eric

[Michael]

I do not find that exact message, but here is what I find that has most of it:

X-Poco-Score-Detail: -100 [%BAYES%=P=50;T=90;BIAS=+20] (%bayes% P=50;T=90;Bias=+20)
X-Poco-Score-Detail: +3 [X-MAILER=] (X-Mailer )
X-Poco-Score-Detail: -4 [TO=%ADDRESSBOOKS%]OR[CC=%ADDRESSBOOKS%] (To %addressbooks%, or Cc %addressbooks%)
X-Poco-Score-Detail: +2 [FROM=%ADDRESSBOOKS%] (From %addressbooks%)
X-Poco-Score-Detail: +100 [SUBJECT=%JUNKSUBJECTS%] (Subject %junksubjects%)
X-Poco-Scored: +1


BTW, when I did the junk mail test, it said it would be considered junk mail under all circumstances. And yes, it is arriving in the In box. (thanks for your help with this - currently my junk mail filtering is only at 20% accuracy.)

Steve

Steve:

The problem here isn't that the phrase isn't being picked up by the junk subjects, it is that the Bayesian filters are thinking the message is legitimate and so are putting a compensating score into the message. Each of the "X-Poco-Score-Detail:" lines represent the results of one of the many checks Poco is doing. A positive score indicates a test that makes Poco think there is an increased likilhood of the message being spam, a negative score indicates that Poco thinks the message is less likely to be spam. I'll go through the scores for the example you posted, these are in the same order as the headers appear in your message:

1. Bayesian filter thinks the message is not spam (you must have set the BF to issue a good score as well as a Junk score)
2. There is no X-Mailer header in the message
3. The message recipients include at least one address that is in one of your address books
4. The message is not from someone whose email address is in any of your address books
5. One or more words from the subject line are in your banned subjects list

What you probably want to do is to decrease your good score in the Bayesian filter settings.

[Bozwell2]

How should I reduce my good score? And to what level? do I change the Good mail bias? do I change the Good Score?

The sender is not in my address list by the way.

thanks again.

Steve

[Michael]

How should I reduce my good score? And to what level? do I change the Good mail bias? do I change the Good Score?

I have my filtering setup with a junk threshold of 90 and a good mail bias of 2.0. I have a junk score of 20 and a good score of 5.

For complete details I suggest checking the "Bayesian Filtering" section of the help file.

The sender is not in my address list by the way.

That's what I had indicated by "The message is not from someone whose email address is in any of your address books", my earlier comment "The message recipients include at least one address that is in one of your address books" was saying that one or more of the "To" or "CC" recipients have email addresses that are in one or more of your address books. This might be your own address if you have put it in your address book.