[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Spam and spam filtering, a problem new to me




On 25-sep-2006, at 22:55, Ron Johnson wrote:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 09/25/06 15:14, Paul E Condon wrote:
Over tha past few weeks, I have started receiving spam email that
contains text the is all well formed words, but doesn't make sense as
a spam message, or as any other sort of communication. I think I have
found what is going on:

The email *does* contain a message. It is contained in a .gif or .png
or other image format file. These are not pictures of naked ladies,
but images of text that touts various penny stocks. If I didn't use
mutt, I might not have had so much puzzlement over them. I suppose
with Outlook all the user sees is the image, which is clearly spam,
but the user doesn't see what the spam filter sees, so, it seems, no
amount of filter fiddling will protect against this. What to do? Are
there new filtering techniques beyond spamassassin?

They are specifically designed to get around SA's traditional
filtering process.  Using sa-learn to train the Bayesian filter,
along with the regular filters, helps a lot.

I've been NOT feeding them to sa-learn, since I thought they would mess up spamassassins bayesian filter. If it trains itself on random non-spam words, couldn't it make the bayesian filter less effective or even start generating more false positives?

Peter



Reply to: