Fog Creek Software
Discussion Board

Solution for spamming

I was struck by a thought today and Im interested in anyone else's opinions.

What if we used server 'tarpits' but with a difference...

if every mail server was configured to (a) recognise spam as it arrived and then (b) reply to that spam automatically (with forged headers, subject etc etc) then spamming would suddenly have a cost for the spammers.
They would have to either manually work through every reply or just give up.

No spammer would last more than 1 or two attempts before giving it up as a bad job.

Thinking aloud
Tuesday, July 1, 2003

They already forge headers, including the email address header.  Some of them operate by distributing the task of spamming across tons of cheap/free dial-up accounts (so even if the email header was valid you'd just be spamming the spammer's mail server -- probably an unwitting ISP).  You could try to deluge them with various network requests if you have their IP addresses (provided that they're valid too), but they could change their sending pattern in such a way that nothing like that could bother them (eg: having multiple dial-up shills and switching frequently).

Tuesday, July 1, 2003

That software exists - it creates a giant network of all users and if enough users delete an e-mail, it will recognize it by the time you check your e-mail and delete it before you see it.

I forget what it's called, but someone here will probably be able to tell you. It's a Google for e-mail.
Tuesday, July 1, 2003

.. I wonder if they've copyrighted the term "SpamRank" lol.
Tuesday, July 1, 2003

"so even if the email header was valid you'd just be spamming the spammer's mail server -- probably an unwitting ISP"

hmm....OTOH the unwitting ISP _does_ need to be informed and this could be a nice way to let them know...

Thinking aloud
Tuesday, July 1, 2003

"That software exists"

<g> the way you describe it sounds totally different.
I was picturing using bayesian techniques to recognise the email as spam, and then automatically have the mail server send a reply to the spammer composed of various random fields.

If the spammer is _not_ a spammer (the email was a false positive) then no real harm done, the sender can see what happened and adjust their wording or whatever.

If it was then the smapper receives umpteen million emails (1 for each email he sent) and has to wade through them all.

If he was routing the email through another mail server then that mail server would receive all the crap....but then it shouldn't be allowing him to route his spam anyway and so it deserves what it gets :)

Thinking aloud
Tuesday, July 1, 2003

That software exists too.
Tuesday, July 1, 2003

Looking at the stuff in my Junk folder, I'd say the spammers don't expect a reply by e-mail but by having you click a link to go to their web site or calling a phone number. So replying back to an e-mail address found in the spam, won't accomplish anything.

Nate Silva
Tuesday, July 1, 2003

My favorite spam revenge has already happened... One of the biggest spammers in the world has been subscribed to nearly every catalog that exists. He gets them by the truckload every day.

My favorite solution today is temporary e-mail addresses. mailshell does this. spamgourmet does this.

I basically create a new e-mail address for everything I sign up for. With spamgourmet I can even define how many times they can e-mail me at that address before they start getting gobbled up. With mailshell, I create rules for the e-mail address, and not the sender.
Tuesday, July 1, 2003

I've found the single most effective filter for spam is to put all HTML email into the sin bin. For me, it catches just about all spam except for the Nigerians, and even they are starting to use HTML now. If I get a false positive, I contact the sender and ask them to send plain text in future - so far, I've found that everyone is more than happy to oblige, and are only using HTML in the first place because that's the default in their email program.

Wednesday, July 2, 2003

Thinking aloud, if you have a good way to "(a) recognise spam as it arrived", then you've solved the problem already.  You just recognise the spam, and quietly delete it.

Foolish Jordan
Wednesday, July 2, 2003

For some spam, we refuse the connection before we even finish receiving it.

This means that our bandwidth costs of receiving the spam are not as high.  Unlike sending out an automated bounce message - which has the effect of doubling your costs...

Wednesday, July 2, 2003

As a business owner, the key issue for me is that the spam blocker can NOT BLOCK any VALID email. 

So, it has to have a 0% false positive rate.

One simple way to do that is the same way they do with Postal mailing lists you purchase on a "per use" basis.

You seed the internet with DECOY email addresses. Any email showing up in those boxes is spam, because those addresses aren't actually used.  If more than one DECOY address gets an email it's certainly spam.

Then you keep a list of the EMAIL (not the sender) and block the email (at the mail server, etc.) that matches the email recieved at the DECOY addresses.

This has been used by paper mailing list folks for years (probably decades) with, I suspect, great success.

Wednesday, July 2, 2003

> Then you keep a list of the EMAIL (not the sender)
> and block the email (at the mail server, etc.)
> that matches the email recieved at the DECOY
> addresses.

Good idea -- in fact, many email services (AOL, Hotmail, etc) do this already.

Unfortunately, spammers are on to that one. All the spammer has to do to defeat this scheme is to append a unique stream of bytes (like a counter of the number of spams sent so far) to the end.


Eric Lippert
Thursday, July 3, 2003

*  Recent Topics

*  Fog Creek Home