Yesterday I saw a spam filter insert a "PROBABLY-SPAM" warning into the subject
line of an innocent message. It also appended what looked like detailed
reasoning to the body of the message: a list of UNDERSCORE_SEPARATED heuristic
names accompanied by a score value. It also said that the overall score spam
threshold was 5; apparently that overall score was the sum of individual scores.
Well, the message was ranked 2.6 by a DEAR_SOMETHING heuristic, and ended up
classified as "spam" (5.3) due to the combined efforts of other incomprehensible
mail-header heuristics. Is it really that bad to start a message with "Dear
Something", and even if it is, is it really more typical of spam messages than
other messages?
Needless to say, the non-technical receiver of the message didn't understand any
of the COMPUTER_GENERATED rubbish at the bottom.
- STUPID_SOMETHING Yossi Kreinin
-
