> couldn't IMAIL's non-Bayesian filtering selectively
> be used in addition to ASSP to catch some of the spam
> that gets through ASSP's filtering?

Short answer: Yes.

Long answer: I ran both ASSP and the Imail anti-spam tests while I was
testing and found the Imail tests unnecessary and have turned them off
completely on my production server to save cycles.  With just the Imail
anti-spam running I was detecting 25-30% of our incoming email as spam with
occasional false positives(mostly opt-in newsletters and the like).  With
ASSP it is too soon to tell on my production server, but on my test system
it was catching 50% of my mail as spam, or roughly 99% of my spam.  I don't
know what to expect for the company's server as I receive allot of valid
email daily due to this and other mailing lists but I also receive more spam
than anybody due to posting on newsgroups, years ago, with my real address
before I knew any better.  So far, over the weekend, it has identified 60%
of our email as spam, but the spam slows down over the weekend as well as
the valid email.

If you have the time to maintain the url blacklist for html emails and the
phrase list for text emails(afaik Imail only compares the email's contents
to one or the other), and can accept that there will be occasional false
positives, they can work together with ASSP but be warned, the lists
provided by Ipswitch have been known to contain phrases and URLs that should
not be blocked.  Some of those URLs have been mentioned on this list.  You
will need to use something to keep track of changes you have made.  I used a
couple of little apps done by Curt Turner that are fantastically helpful:
http://www.turner3d.net/UBMSetup.exe and
http://www.turner3d.net/PLMSetup.exe

During testing Imail's anti-spam, I noticed many plain text spams getting
past Imail.  One thing I did after that, was to copy the contents of the
URL-BL list and merge that with the contents of the phrase list.  This
allowed non-html emails with no banned phrases, but a banned URL, to be
caught.  When I did that, my phrase list matches jumped up 100% as I started
catching many of the plain text emails with bad URLs.

Since ASSP will be the gateway, the Imail DNSBL test will not work, but I
used a RHSBL and did get some hits from it.  It was
"block.rhs.mailpolice.com" and you have to set the type in Imail Admin to
"RHS".  This works by comparing the domain of the from address, I think, so
it didn't matter that the last mail server before Imail was always my own.
This is still a valid test when using ASSP and I may decide to reinstate it
as time goes by.

Allot of valid emails fail some of the HTML content tests, and I believe
every one of them could be failed by a valid email, poorly created, so I
chose not to use them.  If you do, make sure to set the limit high enough to
not catch valid emails, but low enough to be effective.  I don't know what
to recommend for that.  If I was not in a corporate environment, or carried
more clout, I could be more aggressive with not accepting poorly configured
emails like some others on this list choose to do.

And lastly the statistical test is Bayesian like ASSP(but only checks
non-html emails) and by capturing your spam and using the spamseeder
utility, you can build the word list from spam your company receives thus
improving it's accuracy, but if you are using ASSP it would be redundant.

Most of this information is what I have seen in use, or have gleaned from
reading this list, so if I am mistaken on anything, please let me know.

Douglas Traylor
MIS Department
Harbison-Fischer Mfg Co.
(817)297-2211 x210



To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html
List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/
Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/

Reply via email to