Chris St. Pierre wrote:
On Tue, 6 May 2008, mhildebr wrote:
Is there a way to have Spamassassin look for spoofed email addresses
being
used as the sender's address ([EMAIL PROTECTED]) but using a different
user name (Viagra instead of myname)? It seems like it would be
simple to
check the user name and filter results from that. Thanks for any help.
Bad idea. My name can be easily and legitimately displayed in dozens
of different ways, without even considering typos:
Chris St. Pierre
Chris St Pierre
Chris St-Pierre
Chris Saint Pierre
Chris Saint-Pierre
Christopher St. Pierre
...
Christopher A. St. Pierre
...
Chris A. St. Pierre
...
they all match
chris.*pierre
or to be more conservative
[chris]{3}.*[pierre]{3}
And so on and so forth. And if someone accidentally mistypes my name,
they aren't supposed to use your name in their From header, are they?
suddenly I'm Chirs St. Pierre or something like that, and your filter
blocks the message.
A better idea would be to just let Bayes do its thing and notice the
token -- in this case, 'Viagra' -- and score accordingly.
agreed.
Or, you can manually list out the various spellings of each users'
name, and then come up with a fancy algorithm to route around
misspellings, perhaps using something like Levenshtein distance to
figure out how egregiously misspelled a name is. It'll be way more
work than it's worth, but if that's what tickles you, go for it.
