Each test has a corresponding weight. All tests are run against each message; for each test that matches the message, the weight associated with the test is added to the total spam score for the message.
Test scores can be specified as a numerical value (
score), as a percentage (
probability adjustment %), or both. Scores and probability adjustments can be specified to multiple decimal places, although the PureMessage Manager Anti-Spam Rules page only displays three decimal places. Scores and probability percentages can be specified as either positive or negative values; negative values are prefixed with a minus symbol. Positive scores and probability percentages increase the likelihood that a message is classified as spam; negative values decrease the likelihood.
The chart below shows the relationship between scores and percentages. A score of "5" results in a 50% spam probability.
If an anti-spam rule has both a score and a probability percentage, the score is converted to a percentage, and the probability percentage is added to the result. For example, if a rule has a score of 5 and a probability percentage of 10%, the score is converted to a percentage (50%), and the probability percentage is added, resulting in a total score for the rule of 60%.
The PureMessage policy script is configured to perform actions on each message based on the message's total spam score. For example, the default policy adds an
X-PerlMx-Spam header to messages with a spam probability. If the message's spam probability exceeds 50%, it adds the
X-PerlMx-Spam header, and also alters the subject line and copies the message to the quarantine.