Whitelist/verification spam filters

David Mertz, Ph.D. mertz at gnosis.cx
Wed Aug 28 07:49:27 CEST 2002


|I thought that false positive in the context of spam filtering means
|something is falsely classified as spam.  So, by definition, the very narrow
|scenario I described (someone is one your whitelist and they get classified
|as spam) simply cannot happen.

I consider a false positive to be a broader class than you describe.

Suppose a legitimate (but previously unknown) sender sends me a message;
my whitelist/verification filter sends a confirmation challenge; but the
sender's ISP crashes and loses the challenge.  This previously unknown
sender might not be a new -person-, but simply a new address, of course.

I very definitely consider this situation to be a false positive
identification by the system.  It is not the fault of (say) TDMA per
se, but it is exactly the kind of situation that will happen from
time-to-time when someone uses TDMA.  Of course, I don't know what the
actual percentage of false positives is.

The above situation is every bit as bad from a users point-of-view as
the case where SpamAssassin categorized my legitimate message as spam
because my fanciful header "X-Moon-Phase" happened to be attached when
the moon was crescent.  Apparently there is some spam company called
"Crescent", or something with the word in it.  This actually happened to
me.  (Actually, my subject line also contained a question mark... which
was, after all, matched to the subject of the note).

In either case, it is *possible* to manually examine the
"rejected-but-not-yet-deleted" folder... but either way, it's a false
result.

--
---[ to our friends at TLAs (spread the word) ]--------------------------
Echelon North Korea Nazi cracking spy smuggle Columbia fissionable Stego
White Water strategic Clinton Delta Force militia TEMPEST Libya Mossad
---[ Postmodern Enterprises <mertz at gnosis.cx> ]--------------------------





More information about the Python-list mailing list