Hi Ric,

Answered at http://www.logsat.com/spamfilter/forums/showmessage.asp?messageID=3340. We'll figure a way to do this!

Roberto F.
LogSat Software

That was indeed one of our options, but we're trying to improve that a bit to reduce the hit on the database server.

Roberto F.
LogSat Software

Roberto -

"the emails being examined must be *exactly* as they have been received"

That's why I suggested copying all messages that are being sent to the MTA into a database at the same time. The messages will be original, unmodified.  I, as the systems admin, can then develop a web-based interface to sort and select those messages that are SPAM and tag them, and delete the valid messages.  SpamFilter can then see those tagged messages and make the necessary changes in the corpus file so that they'll be more likely to be caught as SPAM by the bayes filter when they come around again.

Again, I understand that this won't work for large ISP's.. but for a relatively small company like mine, it would be great.

-Ric

I think that what Ric is proposing/suggesting is that the messages that are deemed good, get placed in a different quarantine database *prior* to being passed on to the mail server.

This would allow an admin to go back into that database and stamp the message as one that should have been unfit.   No interaction would be needed with anyone's email client.

Ric,

We've been examining various options on how to proceed with this. The greatest hurdle is that in order for the statistical filter to work correctly, the emails being examined must be *exactly* as they have been received. The problem lies in the fact that most users use either Outlook or Outlook Express. These apps, especially the 1st, GREATLY modify the email content, sometimes completely changing the source. If these clients were used to forward the email to a "de-spam" engine, the statistical analysis of these emails would be completely erraneous, and will cause unpredictable results.

The biggest issue is being able to store the original source of the email, to then "de-spam" it if users or admins tag it somehow as false positive. We're evaluating several options and hopefully soon will have something ready.

Roberto F.
LogSat Software

Great minds think alike....

http://www.logsat.com/spamfilter/forums/showmessage.asp?messageID=3126

I have an idea for those of us that are NOT an ISP or large corporation - using spamfilter within our business to reduce costs of dealing with spam, that Logsat may be able to develop to help identify the false negatives slipping by.

Create an option to send all messages that SpamFilter determines to be VALID to a database (as well as delivering them to the intended recipient) so that we can sort through them with a browser interface and manually tag messages as spam (so that the bayesian filter can learn from it's mistakes, like it does now with false positives that are delivered from quarantine) or as not spam.  After a short period of time, the bayesian filter will become so much more efficient, the option could probably be turned off, reducing server overhead.

I understand this wouldn't work for most larger companies or ISP's... the sheer volume of messages would be unwieldy... but for those of us with traffic not exceeding 10K messages a day - it would be functional.

Whatdyathink?

Ric