I'm running spamc/spamd 3.0.2 in Debian.��I�have�Bayesian�tests�turned�on,
and network tests off.

Lately a lot of spam has been getting through to my mailbox.��SA's�false
negative rate used to be about 1%; now it's about 50%.��Looking�at�the
headers for the spam that's getting through, I see that the Bayesian filter
is working correctly: almost all of the spam is tagged as BAYES_95 or
BAYES_99.��My�score�threshold�is�5,�the�BAYES_99�test�alone�(using�its
default value) is worth 4.07, and a few other tests are usually positive as
well.��Yet,�the�total�score�is�around�2.5.��Here's�a�sample�from�today:

X-Spam-Status: No, score=2.7 required=5.0 tests=BAYES_99,HTML_20_30,
�HTML_FONT_INVISIBLE,HTML_IMAGE_ONLY_24,HTML_MESSAGE�autolearn=no�
�version=3.0.2

The scores from the tests listed here should add up to about 5.3, but as you
can see, the total is only 2.7.��So�this�one�gets�through.

I understand that the individual test scores are fed through a neural
network to derive the final score.��So�it�seems�that�this�network�has
started to behave badly.��

Can anyone shed any light on this?��Is�it�a�well-known�problem?��What's�the
preferred way to address it?��Remove�all�of�SA's�learned�information�and
retrain the network?

Thanks,
Andrew.

Reply via email to