here are two main types of circumvent any other barrier you
spam, and they have set up. They have so far, at
different effects on least. But they have to deliver
Internet users. Cancellable their message, whatever it is. If
Usenet spam is a single message we can write software that
sent to 20 or more Usenet recognizes their messages, there
newsgroups. (Through long is no way they can get around
experience, Usenet users have that. Email spam targets
found that any message posted to individual users with direct mail
so many newsgroups is often not messages. Email spam lists are
relevant to most or all of them.) often created by scanning Usenet
Usenet spam is aimed at lurkers, postings, stealing Internet
people who read newsgroups but mailing lists, or searching the
rarely or never post and give Web for addresses. Email spams
their address away. Usenet spam typically cost users money
robs users of the utility of the out-of-pocket to receive. Many
newsgroups by overwhelming them people - anyone with measured
with a barrage of advertising or phone service - read or receive
other irrelevant posts. their mail while the meter is
Furthermore, Usenet spam subverts running, so to speak. Spam costs
the ability of system them additional money. On top of
administrators and owners to that, it costs money for ISPs and
manage the topics they accept on online services to transmit spam,
their systems. and these costs are transmitted
directly to subscribers.
I think it's possible to stop
spam, and that content-based The statistical approach is not
filters are the way to do it. The usually the first one people try
Achilles heel of the spammers is when they write spam filters.
their message. They can Most hackers' first instinct is
to try to write software that what it means. The user doesn't
recognizes individual properties know what it means, but worse
of spam. You look at spams and still, neither does the developer
you think, the gall of these guys of the filter. How many points
to try sending me mail that should an email get for having
begins Dear Friend or has a the word sex in it? A probability
subject line that's all uppercase can of course be mistaken, but
and ends in eight exclamation there is little ambiguity about
points. I can filter out that what it means, or how evidence
stuff with about one line of should be combined to calculate
code. it. Based on my corpus, sex
indicates a .97 probability of
But the real advantage of the the containing email being a
Bayesian approach, of course, is spam, whereas sexy indicates .99
that you know what you're probability. And Bayes' Rule,
measuring. Feature-recognizing equally unambiguous, says that an
filters like SpamAssassin assign email containing both words
a spam score to email. The would, in the (unlikely) absence
Bayesian approach assigns an of any other evidence, have a
actual probability. The problem 99.97% chance of being a spam.
with a score is that no one knows
About the Author:
Find out How Spam Blocker can save you Money at http://SpamBlocker.eask.info
Read more articles by:
Camry James
Article Source: www.iSnare.com
