Difference between revisions of "Spam Filtering"

Revision as of 11:48, 11 May 2009

Spam Filtering

SpamAssassin at XMission

XMission offers a very effective strategy to identify unsolicited email sent to our customers. While XMission has been blocking spam for years, this new spam identification software is catching about 30% of the mail that gets past our already existing spam filters.

The spam detection software is called "SpamAssassin" and it scans incoming email for various characteristics that are often found in spam messages. Each characteristic gets a certain number of points, which are then added together to determine the likelihood that a message is spam. By default, SpamAssassin considers anything with a score of 5 or higher as spam but in our tests some legitimate email can get tagged higher than. This is why we've chosen a slightly higher, but safer, number. The number might change in the future, depending on what our research shows. Know that we will be careful to avoid marking or tossing legitimate email destined for our customers.

The headers that SpamAssassin adds to your email messages looks something like this:

    X-Spam-Status: No, score=6.9 required=8.0 tests=NO_REAL_NAME,PLING,PLING_PLING,HT
    XML_EMBEDS,MAILTO_LINK,CTYPE_JUST_HTML,SUBJ_ALL_CAPS version=2.20
    X-Spam-Level: ******

The X-Spam-Status header shows whether or not the message scored high enough to be considered spam, what the actual score was, what's required for it to be qualified as spam, and what tests/characteristics were positive. The X-Spam-Level header shows the whole number score with asterisks. A message with a score of 6.9 would show 6 asterisks

If the message reaches the score required to be considered spam, the following header is added:

    X-Spam-Flag: YES

You may also choose to filter or sort messages yourself. Unfortunately, most commercial email clients don't support special headers. The most likely options are Eudora 5 or, for more advanced users, procmail.

    Note: Though this is the best spam detection software we've encountered, it is not foolproof. There is a small chance some 
    legitimate email might be incorrectly identified as spam. Use this service at your own risk.