Where did you see that you "Must not" block keywords?  I believe you may have misundersood something there.

Regards,

Dan S.

To Dan,

Yes...that was my my main concern. Having to have keywords in place to use the attachment feature, but to use the Bayesian feature, you have to remove all the keywords for that to work, which breaks the attachment scanning feature. Is that something that will be resolved in the future? or was that by design?

In otherwords, you can't utilize all three methods at once, since the rely upon the different states of the keyword filter either using keywords or not using keywords. I would like to be able to use all three, since that would be the most effective approach in my case.

Chuck

Chuck,

That is actually incorrect. If you can please point to us where you saw that comment we'd like to reply to it with the correction.

You most definetly want to keep your keyword blacklist in place. Any filtering that you have that stops spam should be left as is, this way the Bayesian filter will learn the true picture of your spam.

Roberto F.
LogSat Software

Hi Roberto & Gang,

Ok...I understand what has been said about the keyword situation. Now, with keywords is place, all e-mails that pass all other filters are assigned (is that the right word) a value of 0% for spam. If I remove the keywords, then the Bayesian filter kicks in, and begins to report values such as 47% (example) spam etc. It depends on the email.

While I did this, I noticed at the same time, that spamfilter was not blocking anymore attachments...it was forwarding them straight through. The only way I caught this was two fold. One, I normally see up to hundreds of blocked attachments per day. But when I removed the keywords...this went to zero. Then, my anti-virus on my main e-mail server began catching the infected emails that are normally blocked by the attachment filter.

When I re-inserted the keywords back into SpamFilter, the Bayesian Filter then reports 0% spam on e-mails that had cleared all other filters, and the main e-mail server suddenly was virus-free again. On my install, the Bayesian only works if there are no keywords listed in the blacklist.

I am not sure what's going on, but that's what has been happening on my installation. The only other problems I have encountered are occasional CPU hits at 100%....usually when I attempt to force delivery of mail that I know is not spam to my users. And occasionally SpamFilter crashes...just disappears completely. In this case, All I do is restart the service. There again, it's associated with deleting and delivering quarantined e-mails. Also, I noticed that if I don't run SpamFilter as a service the crashing problem is much less servere, as well as the CPU hits are much less likely to occur.

I run the SpamFilter on Windows 2000 Advanced Server, 1GB Ram on a IBM Netfinity 5000 dual PIII. MDAC ver 2.8 using the Microsoft Access driver.

Roberto...I would be interested in testing that for you....maybe it could help in my situation...thanks. You can e-mail the info to me.

I'll try to duplicate the scenario, at least the logs show the bayesian problem, and is easy to re-create if you need them...just let me know if you need them.

Chuck

Roberto,

Just a heads-up,

I downoaded and installed this new patched version. With the limited testing I have down, all indications appear that the Bayesian filter is now not working at all. Before, if I removed all my kewords, the filter kicked in immedialtely. Now, either way, I don't see the Bayesian filter working at all. In both cases the messages that pass all the the other filter rules, are being reproted as 0% spam.

I have not determined with certainty that the attachment blocking is working or not with the keywords removed. This I plan to do in the morning.

For now, I re-inserted my kewords list and will try again tomorrow when the traffic is much higher.

Was there anything special that I needed to do? All I did was copy and replace the executables within the spamfilter directory. I did stop spamfilter first, and then re-started it after replacing the exe's. Is this okay?

Chuck

Chuck,

Please note that the Bayesian filter is used only if all other tests fail to catch spam. If your regular keywords are very effective, the bayesian analysis will only catch the few emails that are missed by the keyword filter. If you remove the keyword filter and others, then yes, the Bayesian filter will be doing a lot of work.

The upgraded procedure is that simple. Stop SpamFilter, replace the exe's, then restart. That's it.

Roberto F.
LogSat Software