This series of tests perform the Bayesian statistical computations to arrive at the spam probability index called spamicity. By default there are three rules that combine to provide the Bayesian filtering in Praetor. These are explicitly listed for when the ADVANCED filtering mode is set, and implicitly performed if the selection is for the BASIC mode with the pre-configured rule filters. This page explains the three explicit rules that test the Bayesian spamicity value.
|
Purpose: |
Check the Bayesian spamicity to see if it is less than 0.30 (default). |
|
Action: |
Accept if the value is less than this good message threshold level. |
|
Default state: |
Enabled |
|
False Positive: |
No, but it can allow spam to be accepted which represents the false negatives. |
|
Other notes: |
|
|
Purpose: |
Check the Bayesian spamicity to see if it is greater than 0.60 (default). |
|
Action: |
Quarantine if the value is greater than this bad message (spam) threshold level. |
|
Default state: |
Enabled |
|
False Positive: |
Low incidence of non-spam messages being rated as spam. |
|
Other notes: |
Once you feel confident that this 0.60 threshold has no visible false positives, you may want to change the action to reject. |
|
Purpose: |
Check the Bayesian spamicity to see if it is in the range between 0.30 and 0.60 (default). |
|
Action: |
Accept if the value is in this unsure range. |
|
Default state: |
Enabled |
|
False Positive: |
No false positives, but likely to be the source of some false negatives which are spam that is not caught. |
|
Other notes: |
If the threshold values are adjusted to your satisfaction, it may be better to change the action to Quarantine and deal with the fewer false positives. |