Configuring an e-mail client to filter spam

If you read your DU e-mail with webCentral, pine on Agora, or another mail client that connects to the central mail servers, we recommend using the Email Account Manager to configure a mail filter.

Please note: The College of Law, Computer Science, Psychology, Physics, and the Women's Foundation of Colorado each maintain their own mail servers that receive e-mail directly from the Internet. Thus, e-mail from the Internet to these systems bypasses the central spam mitigation measures. Please contact the appropriate support staff in each area if you have any questions concerning spam filtering for accounts on their servers. This includes messages accepted by one of these systems and forwarded to the central e-mail servers.

If you read your e-mail on one of the departmental or college e-mail servers, or forward your e-mail to an off-campus site, your e-mail may not be filtered for spam. If you receive messages tagged by PureMessage, and your e-mail client is capable of filtering, you may wish to configure it to do some filtering for you.

Not all e-mail clients can be configured to filter spam. If you are interested in having yours do so, this page gives basic information about the anti-spam headers added by the central mail servers at DU. For information about how to use this information to filter spam from your Inbox please consult the Help section of your e-mail client software.

If it is possible, have the filters look for an e-mail header of the
following format:

X-Dispatcher: DU-Spam-Warning: %%%%% PureMessage reports: Spam-Status: Yes

In addition to PureMessage we use the information from various spam block lists to add similar headers. The lists we use contain the IP addresses of the origin of messages individuals or processes have identified as spam. The number of %'s indicates how reliable we believe the particular block list to be in accurately identifying known spam sources. The number of %'s varies from 1 to 5 with 1 being the least reliable and 5 the most reliable. The score of 5 %'s is reserved for PureMessage.

Since all of the X-Dispatcher: DU-Spam-Warning headers contain %'s, and some explanatory text, it is possible to chose how aggressively you wish to filter your e-mail by using the number of %'s as a scale of "spamishness". Having your filter configured to look for a header such as this:

X-Dispatcher: DU-Spam-Warning: %

Will result in aggressive filtering.

While having the filter look for more %'s will make it less aggressive. This would be the least aggressive:

X-Dispatcher: DU-Spam-Warning: %%%%%

In addition, PureMessage adds its own headers to messages. For an email that PureMessage considers likely to be spam (because they have "spam scores" of 5 or more) the headers look like this:

X-PerlMx-Spam: Gauge=XXXXXXXXXIIIIIIIII, Probability=99%,
Report='URI_CLASS_OTHER_DOMAIN 8, IMGSPAM_BODY 3.2, RCVD_IN_CBL 3,
CHARSET_FARAWAY_HEADERS 1.154, __CT 0, __CTYPE_HAS_BOUNDARY 0,
__CTYPE_MULTIPART 0, __EMBEDDED_IMG 0, __EXTRA_MPART_TYPE_N1 0,
__HAS_MSGID 0,
__HAS_MSMAIL_PRI 0, __HAS_X_MAILER 0, __HAS_X_PRIORITY 0,
__HTML_FONT_GREEN 0,
__HTML_FONT_RED 0, __IMGSPAM_BODY 0, __MIME_HTML 0, __MIME_VERSION 0,
__RAL_ROLEX_BRAND_1 0, __SANE_MSGID 0, __STOCK_PHRASE_7 0,
__TAG_EXISTS_HTML 0, __cbl.abuseat.org_TIMEOUT ,
__dnsbl.njabl.org_TIMEOUT ,
__sbl.spamhaus.org_TIMEOUT '
X-PMX-Version: 5.0.3.165339, Antispam-Engine: 2.1.0.0,
Antispam-Data: 2005.8.24.11

When messages have a spam score less than 5, the headers look like this:

X-PerlMx-Spam: Gauge=IIII, Probability=5%, Report='__CT 0,
__CTYPE_HAS_BOUNDARY 0, __CTYPE_MULTIPART 0, __HAS_MSGID 0,
__MIME_TEXT_ONLY 0, __MIME_VERSION 0, __SANE_MSGID 0, __USER_AGENT 0,
__cbl.abuseat.org_TIMEOUT , __dnsbl.njabl.org_TIMEOUT ,
__sbl.spamhaus.org_TIMEOUT '
X-PMX-Version: 5.0.3.165339, Antispam-Engine: 2.1.0.0,
Antispam-Data: 2005.9.3.8

One could choose to use the Gauge= in a manner similar to the X-Dispatcher: DU-Spam-Warning: header. It is a graphic representation of the score PureMessage assigs to the message, in loose Roman Numeral format. The above example has a probability of %5; if Gauge were to read "Gauge=XXXXXXXXXIIIIIIIII", the probability would be %99.

Whatever the filtering capabilities of your e-mail client, we urge you to validate that the rules you are using are effective and that they do not inadvertently discard messages that you might wish to read.

Please note that this is informational only and does not imply that UTS will support custom filtering configurations. We also reserve the right to change the headers or content of headers added by PureMessage or similar mechanisms.