re: spam filters...
re: spam filters...
- Subject: re: spam filters...
- From: David Crowe <email@hidden>
- Date: Tue, 8 Apr 2003 13:13:55 -0600
I use Eudora and have extensive filters for spam. Eudora encourages a
craftsmanlike approach. It doesn't automate the process, but the
results are probably better than other approaches, if you don't mind
putting some thought into the problem.
My strategies are to focus on the things that spammers can't change,
or the things that, in trying to be invisible, actually reveal them:
Words in the subject line (this gets only the amateurs). Such as
"This is not spam" (okay, really stupid, but my last hit on this one
was only about a week ago).
Stupid HTML (note that Eudora filters the HTML source), such as
">a<!--" (i.e. a single letter between two comments, used to break up
the HTML to prevent anti-spam scanning techniques, while leaving a
word like Viagra on the screen, but something like
"<!--xyz>V<!--xyz>iagra visible to the casual reader).
Character sets that you can't read, such as: content="zh-tw" (this
means that the message is in Chinese, for example, and is usually
spam ... not the case for people who speak Chinese!)
Funny characters, such as capital Omega (probably indicates that the
spam is poorly formatted or in a foreign language). Again,
mathematicians might want to avoid this.
URL's embedded in the body of the email, such as "/www.erotic." or
"
http://optin";. I only filter the minimum recognizable piece of the
URL. Spammers can't change these too much or nobody could respond to
the spam! So, it's a real weak-point.
A name like yours that is in a cc list (yes, some spammers are
stupid). This works because they usually go through lists
alphabetically. So, if I see "crowbar" in my cc list, it's spam.
Phony removal text, such as any mention of S.1618, Senate Bill 1618,
H.R. 3113 or HR 4678, or promises to be removed "to be removed from
our mail". Unfortunately, you need lots of these.
References to the Nigerian scam (e.g. the name of any recent African
dictator, "Urgent/Confidential business proposal).
The "From" portion of the email can be scanned, not for the actual
email address, but sometimes the accompanying name contains keywords
like Septic, GossipFlash, Chivas Regal etc.
Doing this is a bit of work, but my rate of false positives (valid
emails identified as spam) is very low, probably only about 1% of
what goes in my spam box. False negatives are higher, probably 5% of
spam doesn't get detected. But, over time I do see an improvement.
Think of spam like deep sea animals. These animals are generally
transparent, but few are invisible because eyes cannot be invisible
(they have to reflect light). Spammers are the same, if they don't
have a URL in their mail (or a 1-800 number) how is anyone going to
contact them?
- David Crowe
_______________________________________________
applescript-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/applescript-users
Do not post admin requests to the list. They will be ignored.