Re: Some last info regarding analysis

Jose Marcio,

........
>But when you visually look at it, you can near allways class it in a 
>spam/not spam.
>
>- you do that based on what you learn before about spam - this means
>   spam isn't volatile
>- there are too much redundancy on spam - you can identify that
>   based on non-volatility.
>- both of above options - I think this is true.
>
>I'm thinking about how can we think about spam sources and create
>a model comparing them with information sources.

If we had a workable process of visual inspection, it would probably cut 
more spam, faster than any other method (with the spams we have today), 
problem is how to define and describe the pure structural topology of a spam 
and what happends when the spammers adopt?

The issue of languages also still exist. A valid letter in greenlandic or 
inukit(Canada) would initially look like spam for an english-speaking (as 
spammers nonsens words). Do we need spamfilters that is language specific? 
How to chose the right one?

And since you bring up spam sources, from my own work, the phone numbers and 
URLs is still an important source of identification. After some months with 
a low number of spams in the mailbox, 3 weeks ago I hade a surge (from 1-2 a 
day to 10-15 of a steady 30-35 spams/day) and suspected a design flaw, but 
after analysing the spams it was a number of new domains. After extracting 
ca 10-15 new domains, I was back to my original figures.

Regards Kurt

_________________________________________________________________
MSN 8 helps eliminate e-mail viruses. Get 2 months FREE*. 
http://join.msn.com/?page=features/virus