|
Hey All,
I’ve been using Qmail-Scanner for about 3 months now on our qmail-ldap cluster with great success – we get zero viruses through and are tagging %60-%70 percent of our spam with almost no false positives.
I would like to start using the bayes filtering included with spam assassin and we are talking about developing an add on to qmail-scanner to allow people in our company to simply forward a message to an address to flag it as spam or not spam (bayes-spam@domain or bayes-notspam@domain or something of the like).
I won’t go into too many details about the architecture as it is still in the planning stage, but the goal is to be able to pull the original message from a database based on the message that gets forwarded to a particular address.
Is anyone interested in such a technology? What are the positives and negatives you guys see with this? Is there anyone already working on such a beast?
To short circuit a few gotchas you might be thinking of, we are planning on implementing a ratio system to prevent more spam from being analyzing in the corpus than not spam. Of course, there would be access control so that not anyone could inject messages into the above addresses. We are also planning on using a system of crontabs to inject the messages into a database with various MACing mechanisms to help restore the original message with full headers.
What do you guys think? |
