|
Easily Trainable
Advanced Bayesian Filter |
Unlike other competing products, Praetor's implementation of the
advanced spam detection technique known as Bayesian Statistical Filtering is trainable. The administrator can train the Bayesian filter
in Praetor either "on-the-fly" individually upon
reviewing the message details or in batches of several messages.
|
|
IMMEDIATE TRAINING
|
The administrator can train Praetor by reviewing a message in
Praetor's Log Viewer. Based upon the content of the message, the
administrator determines if it is spam (unwanted email) or ham (good
email). Pressing the appropriate button ("Train As BAD" or "Train
As GOOD") immediately trains Praetor's
Bayesian filter.
This means Praetor can filter more accurately with fewer false positives when
trained with samples of the
good and spam messages that your site receives. (For those
interested, our customers report that the default token database will
successfully filter greater than 90% of the spam arriving at your site,
with a false positive rate less than 0.5%.)
|
|
BATCH TRAINING
|
To facilitate training, Praetor has an optional means to capture all
messages classified as ‘uncertain’ by the Bayesian filter. This allows the administrator
to review and train multiple messages in a batch mode.
Praetor also simplifies the accumulation of message samples for on-going
Bayesian training. Any messages tagged as unsure by the Bayesian
filter can be automatically saved for further training by checking the
"Save UNSURE" box.
Periodically the administrator would review
Praetor's Train Unsure Folder and determine if each of the captured
samples are spam or ham.
 |
|
TRAIN UNSURE FOLDER
|
|
Messages designated as "UNSURE" by Praetor are held in the
"Train UNSURE" folder. The administrator can
periodically examine the contents of this folder and create
Bayesian training queues based upon whether these messages
are spam or ham. |
BL-Monitor
