spf-discuss
[Top] [All Lists]

Re: Spam undetectable by SPF (domain reputation) or Bayesian (content)?

2004-09-10 07:39:13
However, the point is that spammers do send e-mails that are 
entirely random text content, so they could learn to send randomized madlibs 
instead.

This is a special case of a more general phenomenon.  Spammers have been 
forced, by the more effective antispam methods, to adopt practices that 
severely limit their business success.

Everbody agrees that the most ideal way to stop spam would be to educate all 
the world's users to never ever buy something from a spammer.  Many of the 
effective antispam schemes have the effect of forcing the spammers to do this 
education themselves.  Disorganized, semirandom email messages do not resemble 
real advertising or inspire much confidence in a potential customer.  When a 
spammer has to spell his product  v*1(_at_)g R0  , he is pretty severely 
limited, and his potential customers are given a big clue that the seller may 
not be legit.

No legitimate business today will advertise with a spammer.  This is itself a 
victory for antispam efforts.  It was not at all clear, in the earlier days of 
spam, that legitimate business might not adopt similar spam practices.  Now 
they realize that, as an advertising medium, spam is hoplessly polluted and 
limited.  Companies that run legitimate email advertising today are 
scrupulously careful to do things right.  (At least the ones I have allowed to 
continue sending me advertising.)

In short, I don't think anti spam measures need to be 100% effective on their 
own to have long term effectiveness.  Every time we raise the bar for spammers, 
we reduce their ability to make money from their efforts.  Forcing them to buy 
throwaway domains means they either have to spend some of their own money, or 
steal someone elses.  The former cuts into profits, the latter exposes them to 
conventional law enforcement.  Forcing them to buy throwaway domains also 
forces them into at least a psuedo legitimate business relationship with the 
domain registrar.  That opens another avenue of attack against them.

Remember, spammers adopted domain spoofing in response to outside pressure.  It 
was not a "natural" choice for them.  They could have chosen throwaway domains 
instead, and in fact did for a while.  They chose domain spoofing because it 
was easier and/or less expensive for them and exposed them less.  If we take it 
away, we are forcing them to give up a choice they made for their own 
advantage.  We are driving them backwards.  That hurts them.

The all or nothing argument is being used against SPF and other authentication 
schemes in the press right now.  The argument goes: SPF doesn't stop all spam, 
for all time, everywhere and therefore it is useless and should be abandoned.

When my son took a logic course a couple of years ago, he learned the latin 
name for this type of fallacy. (I don't recall it now.)  It is such an old kind 
of false reasoning that it has a name in an antique language.

Mark Holm