[NMIA Logo]
New Mexico Internet Access
 
Home
|
NMIA's services
|
Manage Your Account
|
Support
|
Message of the Day
|
Check email via the web

Click here for a message from NMIA about spam or here about why filtering is difficult or here for a RECIPE for reducing spam or here about AOL blocking email.

About Spam Assassin

NMIA email now incorporates a system called Spam Assassin (SA). By itself, this system does not delete spam. It does subject each email to a suite of tests, each with an associated score (positive or negative), and reports the algebraic sum as an index of the probability the message is spam. A higher number means the message is more likely to be spam. The score is a computer's guess and can be way off. But the score can serve as a useful filtering criterion (see Recipe below)

This score is recorded on the "X-Spam-Level:" header line. Most headers are hidden by email client programs (Outlook, Eudora, etc), however we can make the score visible, if you request it. It will appear as a number of capital Q characters on the "Subject:" line of each message. More Qs means it is more likely spam.

This will allow you to watch the scores which SA reports on desired emails (ham), and use them in the RECIPE below. Just send an email to help@nmia.com to have this Spam Assassin score appear on the "Subject:" line of your email.

The RECIPE can help you determine the score that surely means spam, and make the filtering completely automatic on our systems. Once you have determined that score, let us know and we can eliminate it before you see it.

Most of our users have decided to set their SA score filter at 4, 5, or 6 Qs, but you must use the RECIPE to determine it. This can be adjusted as experience dictates.

-- help@nmia.com

RECIPE for Using Spam Assassin


This is a recipe for using Spam Assassin scores to mitigate spam before download.

For some time now, we have been using a computer program to scan all email passing through our systems. It is called SpamAssassin (SA), and it examines email headers and content for tokens known to be associated with spam. It provides a score, a number, included in a special header line, which is a statistical guess at the probability that the message is spam. The scoring is not perfect, and sometimes gives inappropriate scores, but millions are using it now.

If properly and carefully followed, this recipe will reduce received spam levels dramatically. It takes patience and a little time to work properly.

These steps _must_ be taken in sequence, to best manage spam.

1.) Ask to have the SA scores added to the "Subject:" line, where you can see them (normally the score is not seen by most email handlers, i.e., Outlook, Eudora, Netscape, etc). You will begin to see the score as a series of capital Q characters at the end of the "Subject:" line of any message given a positive score by SA. Negative scores are not shown.

2.) (Very Important) Ignore all spam messages and their scores. For a week or two, watch ONLY the ham scores, i.e., on desired messages. Make note carefully of the highest SA score you see on any identified ham messages you get. The Spam scores are not important, they may be very high or even zero, but we must know the highest score seen on _your_ ham messages.

3.) After a few weeks, let us know that maximum score and we can set a discard filter a little higher than that value. Thus you will preserve all ham and eliminate much of the spam. Note, the scores on spam may be higher or lower than that value, no matter. After this is done;

4.) If it seems appropriate, you can then collect samples of any ham messages which you regularly receive, which have the high scores. (only regularly received ham applies here, e.g., mail lists, relatives, friends) Call us and we'll explain how you can forward the headers of those messages to us so that ...;

5.) We can make a white list filter which will flag such messages to bypass any filtering based on SA scores, ensuring delivery of those desired emails. This, in turn, will allow us to lower the score used in the discard filter, getting rid of more spam. This can be done progressively one Q at a time.

6.) This next step may not be needed. If a few especially offensive and repetitive spam messages become annoying, their header information can be collected and used for a black list filter to force their discard. Unfortunately, variations on words are nearly impossible to predict. Therefore, most black lists do little good since the offense rarely repeats exactly.

Other than fine tuning each of these steps, there are a few heroic measures NMIA can consider if you are still getting much more spam than ham.

It cannot be over emphasized: The scores you see on spam messages are of no consequence. If it's spam it's spam, no matter the score. What matters is the scores you find on messages you want to receive, i.e., on ham. So begin by ignoring spam; and focus on the scores assigned to your ham messages. They are the key, and necessary to tuning the best discard filter. That, in combination with a possible white list does most of the job.

Please let us know if you have any questions or when you are ready to begin the next step in the process. Help Desk at NMIA

If You Want To Do It On Your Computer

If you would prefer not to have SA score filtering performed on our systems, you can find instructions for setting up filtering in your mail program here:

http://www.nmia.com/spamfiltering.html

This page includes instructions for setting up the following email clients:

Eudora 4.x
Eudora 5.x
Outlook Express
Outlook 97
Outlook 98
Outlook 2000
Outlook 2002
Netscape 4.7x
Netscape 6.x
Netscape 7.x

If you have questions, please don't hesitate to contact us: help@nmia.com

A message from NMIA about spam:

Please be assured that NMIA is very active in addressing the spam problem. Our subscribers names, email addresses, and all other personal information is scrupulously kept private, and is never offered to anyone for any purpose whatsoever. We do not spam our subscribers, and we have no tolerance for any subscriber who attempts to spam others. We maintain extensive lists of IP addresses from which spam has arrived, and block all such until the owner has been proven innocent of the offense.

In the last few months and weeks, the flux of spam for everyone has increased dramatically. Were it not for careful filtering, some of our subscribers would receive in the order of a thousand spam messages a day. With filtering, in many circumstances this can be reduced to 10 or 20 a day actually being downloaded to the subscriber's computer.

Any email address that has become spam exposed, will usually suffer a burden of spam, increasing more or less continually and indefinitely. This is the nature of the Internet today.

An email address can be exposed by any of a number of events. If the email address appears on a web page, (almost anywhere on a web page) it will be exposed within a month or so by spiders randomly gleaning such information. The same is true if the address is merely used in the domain registration, without ever creating a web presence.

Answering any spam message, even to request deletion from the list, is a nearly certain way to confirm that the address is active and to more widely advertise it. Any correspondence with businesses that partner with consumer habits data collection services is likely to result in spam exposure. Filling out information request forms on web sites, or to "join" anything, should be routinely spoofed by using a false address or possibly an alias to one's real address and then checked to see if it has been spam exposed.

One recent and very serious exposure mode occurs when the computer, of a person with whom you have had any email correspondence, becomes infected by any of several viruses (the Sobig family, e.g.). Such viruses gather the email addresses of anyone who has sent email to, or received email from, the virus victim. These address will appear in large collections of addresses sold to spammers, within a few days or weeks. The person infected, of course, is also exposed.

Some otherwise legitimate Internet online merchants provide their email contact lists to "partners" who often are not as honorable.

Just being active on the Internet is an easy avenue to spam exposure, e.g., newsgroup posts, some blog activity, publications that rate references on related sites, even posts on some mailing lists, etc. One widely advertised and inexpensive anti-spam program offered on a web site turned out to be an address collection scam. Recently, a great deal of spam is sent to addresses that are randomly generated with combinations of letters sometimes combined with first letters of both proper and given names found in telephone directories. Much skilled programming expertise is wasted in providing automated systems for such efforts.

In general, it is not possible to purge the Internet of one's address once it has been spam exposed.

The only remedies are extreme filtering or a change of email address. Filtering is effective, but can be tedious to optimize. One of our staff, over a period of several weeks, having carefully categorized the Spam Assassin scores of all their email and tightened the SA filter to the point of losing some ham (i.e., desired email), was able to design a filter based on SA but including a short list of "unconditional allow" addresses, which reduced the flux of spam from 60 ham messages and 980 (typical) spam messages to 60 ham and about 50 spam, most easily discerned and deleted manually. This not for the impatient or easily discouraged.

There are other basic filter programs that are approximately as effective, but spammers learn to bypass the filters with their message creating programs. Therefore a robust Bayesian filter program (such as SA) is a wise starting point.

Changing addresses, if it can be accomplished without too much pain, can be effective, with certain caveats. The down side of an address change varies greatly with how the email address in question is used, i.e., business impact, loss of remote and occasional correspondents, how new address information can be distributed, etc. In many circumstances the required advertisement of the new address will cause the relief to be short lived. In some such cases, advertising an alias for the real address may help.

NMIA is, of course, available for email or in-office discussion and suggestions, and for any assistance in filtering or address changes. There are no fees associated with basic filtering, nor with any manipulation of addresses or aliases. Please know that we are very interested in, and competent to address, solutions to this almost ridiculous problem.

Please complain to your elected representatives, who have created a spam control bill which will have zero effect on spam, and who have also flatly rejected all International overtures to cooperatively address this problem.


Maintained by:
www@nmia.com