[Ham-Computers] RE: Spam Filters

[Hsu, Aaron (NBC Universal)]

Yep.  And, in many respects, Bayesian filtering might actually work better with non sensical phrases.  You'll notice that many of these types of e-mail contain syntatically correct phrases, but the verbage is basically garbage.  The more times a certain word or phrase occurs in SPAM, the better the Bayesian filter is able to determine that these word groupings point to SPAM.  The "better" Bayesian SPAM filters out there catch 90% of SPAM of which most which are not caught are subsequently marked as "Likely" or "Possible" SPAM.

Note: due to the way Bayesian filtering works, the more e-mails samples it has to create a database, the better it is at sorting bad from good.  In other words, the more e-mails you get, the better it works.  BTW, both bad and good e-mails are used to determine if an e-mail is considered legit or SPAM.  And, most Bayesian filters also allow black and white lists to futher enhance filtering.

If you use the "full" version of Outlook (not Outlook Express), I highly suggest you take a look at "Inboxer".  I've used it for a while and it really does filter 90+% of the SPAM going to my personal e-mail addresses.  They also have a hardware appliance that will work on a network.  How well does it work?  In the course of a week, I get about 1000 e-mails to my personal e-mail addresses.  On average, about 25 are legit and about 50 are marked as "likely" - the rest are dumped into the SPAM folder.  In the SPAM folder, I browse through the subject lines of the ones that are addresssed to me by name or my e-mail address - the rest get the delete key.


  - Aaron, NN6O


-----Original Message-----
Sent: Monday, April 16, 2007 10:50 AM
Subject: [Ham-Computers] Spam Filters

Aaron, is that why these days there is so much spam consisting of a long series of unrelated, often 'non-sensical' words?

John
WB6BLV at inreach.com