Print

All messages handled by the AcceptIO servers go through a scoring process to rate the likelihood that they are spam. Higher scores mean a higher likelihood of the message being spam. Theoretically, there is no limit to how high or low a message might score. In practice, virtually all messages scoring from 5 to 40 are definitely spam by anyone's standard. Messages scoring below 0 are pretty likely to not be spam. Messages scoring from 0 to 5 should really be looked at by a person instead of a program to make a final decision.

From time to time, AcceptIO publishes statistics on the spam scoring process. We recommend that you look at a sample of the actual spam scores on messages you receive and come to your own conclusions about what ranges of scores make sense for your messages.

The spam scores and other details are placed in custom email header fields by the AcceptIO service. You probably don't see these by default, but you should be able to get to them somehow. The actual method varies by email program, so look for an option like "view raw message" or "view message source" or "view all headers". Here is an example from a recent actual message:

X-SA-Spam-Score: 6.8
X-SA-Spam-Score-Graph: ++++++
X-SA-Spam-Report: SpamAssassin 3.3.1 running at AcceptIO.com,
scored 6.8 on Sat, 12 Nov 2011 15:39:55 -0800.
details: BAYES_50=0.8, EM_ROLEX=0.618, FREEMAIL_FROM=0.001, RDNS_NONE=0.793, TO_NO_BRKTS_NORDNS=0.001, URIBL_DBL_SPAM=1.7, URIBL_JP_SURBL=1.25, URIBL_WS_SURBL=1.608

You may be able to configure your email program to automatically act on messages based on what it sees in those email headers (even if they are not visible to you). The headers have the following meanings:

X-SA-Spam-Score: Gives the numeric score that came out of spam scoring. It can also be negative. A higher score means the message is more likely to be spam. If your email program can do filtering rules based on numeric values in headers, then this is the best header to use.

X-SA-Spam-Score-Graph: This is equivalent to the numeric score, but it's expressed as a series of plus signs (for positive scores) or minus signs (for negative scores). If your email program can only do filtering based on text strings, then you should use this header.

X-SA-Spam-Report: This header gives details of what the spam scoring process found. It is difficult to use this in an email filtering rule, but it is sometimes useful in figuring out why a message was misclassified.

In addition to things you can do locally with your email program, the AcceptIO service has three preference settings related to spam scoring for every mailbox. These values are compared to spam scores for messages and can be set on a per-mailbox basis.

  1. Reject message. The best way for our servers to handle spam is to refuse to accept it at the time a calling server tries to give it to us. When we do that, we give the calling server a notification that we are not accepting the message. This is best because the sender of a good message -- that happens to score too high -- will get a notification that their message was not accepted. Because of some technical characteristics of the SMTP protocol and best practices, that's not always possible for us to do. We can do it if all of the recipients of the message agree that it should be rejected, as reflected in their reject message preferences. If any recipient doesn't agree (or if any recipient is a non-AcceptIO address), then we cannot reject the message this way. Suppose a message with two recipients gets a spam score of 14.7. One recipient has a reject message preference of 5.0 and the other has a reject message preference of 10.5. Since both recipients are in agreement that the message should be rejected, that's what will happen. The default value is 50.0. That's a fairly conservative default value since our philosophy is to not block messages without your involvement. You might consider making it something more aggressive (we suggest 10.0).
  2. Silently discard message. Messages with spam scores higher than this will be silently discarded and not delivered to your mailbox. This is a fairly drastic action because the originator will not receive any notification that you didn't get the message. You should be comfortable with that idea before changing this value. The default value is 99.9.
  3. File message to Spam folder. Messages scoring higher than this will be filed in your Spam folder instead of your main mailbox. The idea is to pick a value for this preference where messages are pretty likely to be spam. Most people like to check their Spam folder periodically to make sure no good messages have been placed there by accident. (If you don't see a Spam folder, you may have to tell your email program you would like to subscribe to it.) The default value is 5.0.

How should you think about setting values for these preferences? We like to think of them in terms of the risk of lost messages. The reject message score is the least risky since the sender of a good message should get a notice that the message was rejected. The file message to Spam folder score is reasonably safe if you periodically check your Spam folder for good messages. The silently discard message option is the riskiest, and you should have this set to a fairly high score (or just use the system default value). With the defaults mentioned above, messages would be processed as follows:


NOTE: The user interface for adjusting these AcceptIO service preference values yourself is not yet available, but the spam scoring feature is operational. If you would like to set values for any of them (or would like to see what your current values are), please contact AcceptIO support.