Dealing with spam within DCIS

Dealing with spam within DCIS

Spam is an ever-growing problem for email users within DCIS. To keep the problem manageable, we use SpamAssassin, which uses a large number of rules and a Bayesian filter to score the likelihood that each message is or is not spam. Headers are added to email before it's delivered with the results of the tests.

No matter how highly email may rate on these tests, no email is ever deleted by these filters before you get to make the final decision. Filters are not perfect and will make mistakes. Some legitimate mail will be tagged as spam (false positives) while some spam will not be recognized (false negatives).

Since the spam filters are not perfect, you will have to look at the spam at some point. You should ignore your spam folder only as long as you would want an important email delayed. The advantage of tagging the spam is that you get to choose when you have to look at it.

Suspected spam can be filtered into a separate folder via a variety of methods. If you read your email on, you can set up rules to do it there. Mail addressed to most clusters within DCIS is now tagged automatically if it appears to be spam. You can tell if it is by looking for a line like

    X-Spam-Status-LSCR: dragon spam scanned
If you read your mail on a unix machine, you can use procmail to put it in a different folder. If your mail is not automatically scanned/tagged for spam, you have to use procmail to both pass it through SpamAssassin and file it based on the results.

Note: Since the new Bayesian filter learns what spam looks like based on training, it it now possible to give feedback to the filter to improve its performance.

This page last updated July 5, 2005.