Re: Inline image spams

Darxus-5+gNnNaL/BD/[email protected] Wed, 20 Dec 2006 07:41:11 -0500
Newsgroups gmane.mail.spam.spamprobe.general
Message-ID <[email protected]>
On 12/20, Jem wrote:
> images. Might normal (non-spammer) people elsewhere put inline images into 
> emails? I just don't know enough about this.

Yes.

> Glad you point that out, this is true. I certainly would not worry about 
> such a proposed change, the only way it could cause a false positive would 
> be a totally new contact deciding to send me images in the email. But 
> generally we call those spammers, so ... yeah I like the idea.

I'd rather have the false positives from email from a new legit
sender result in a bounce (error message back) to that legit sender
from my postfix body_checks instead of ending up in a folder full of
spam that's never checked: http://www.chaosreigns.com/spam/#image

> very tempted to try writing a simple bitmap-image analysis library that 
> could quantify some aspect of an image. An easy one would be edge or line 
> detection which should differentiate between an image heavy in regular 
> shapes like text, versus a photographic image.

You belong on this list:
http://lists.own-hero.net/mailman/listinfo/devel-spam
They are actively developing similar things, in addition to having some
success using OCR.

> Of course this would push the spammers to start creating fuzzy text and 
> other unusual graphics, which honestly would bring me a bit of pleasure 

They're already doing it.  And I think your expectation that it will result
in fewer people paying attention to them as a result is incorrect.  They're
taking advantage of their need to obfuscate the text to make it look
more... festive.  I'm actually impressed at how well the OCR folks are
handling it, but I don't think it's really feasible to keep production
level OCR based image spam filtering ahead of the spammers.  

> It would be exciting to have something like that which could become a score 
> input to spamprobe or other software. I've done image analysis before in 
> computer vision courses but it was all with Matlab, have never coded it in 
> C before. But in principle this could be quite simple, iterative, linear.

I continue to be convinced that the only solution to image spam is to block
all emails with inlined attached images, rejecting them before the end of
the initial SMTP session so that the sender is the one to send any bounce
messages, so spammers don't send bounces (backscatter) and legit senders
do get an error message.

However, I would like to be proved wrong.  Get to it :)

-- 
"Every normal man must be tempted at times to spit upon his hands,
hoist the black flag, and begin slitting throats."
 - Henry Louis Mencken (1880-1956)
http://www.ChaosReigns.com

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV