Hi Kyle,

good questions!  Sorry for the length in
answers.


> I've been looking at the GeoTrust papers, and I'm struck by two things:
>
> 1) There are multiple CAs with 'authority' (granted by the Mozilla
> Foundation, by including their root certificates with the
> distribution) to issue certificates, and they do not coordinate their
> actions;


Yes, Mozilla acts as a super-CA for the
initial distribution.  Now, one might think
that this means it should adopt the same
care as the CAs ... but along side
this is the fact that anyone can distro
the product, and in fact many do, and
change the root list as they desire.  So
that says no care should be taken...  Add
to that the processes of other browser
manufacturers:  Opera takes money.  M$
requires a WebTrust or equivalent national.
Konqueror does whatever Mozilla does.

To bring some semblance of order to this
chaos is the Frank Hecker policy which has
now been put before "staff".  The gaping
hole of what it all means desperately needs
to be filled, for reasons I won't go into
now.


> 2) each of the cases that GeoTrust asserts as failing examples relies
> on a trademark infringement.


One caveat.  I would suspect that the GeoTrust
people modelled the attacks along the lines of
trademark infringements because that's what they
understood.  In contrast, the original SSL
architects modelled the structure based on old
crypto threat models taken from the military,
because that's what they had.  Meanwhile the
browser implementors modelled it on user gripes.

And phishers model it on money received.  Just
because one group says "it is X" doesn't mean it
is, it often means that X is all they are trained
to perceive.

> The first thing is a contributor to all of this trouble.  There's no
> "industry best practices" that the industry has had the chance to go
> over and refine; instead, they're all working from the
> ANSI/WebTrust/standards-body ideas of how CAs should behave, without
> revision and without oversight.  I'm surprised it took as long as it
> did to have problems.


Right, what I mentioned above.  Bear in mind also
that the system was put in place for a *perceived*
threat, not a validated threat, so all of what you
call "best practices" was written without direct
reference to any validated threats (hence the
dramatic difference between GeoTrust's statements
and other practices).  But see more on that below.

> The second thing, though, is a legal argument, and one that deserves
> more rational analysis.
>
> In the US (at least; I don't know anything about international law as
> relates to trademarks, tradenames, et al), banks and insurance
> companies file their trademarks under Class 36, and telecoms file
> their trademarks under Class 38.  These are the two (along with Class
> 45, "personal services", if "providing a credit report" is a function
> of a personal service) that seem to be most likely subject to phishing
> attacks.  (Though, to be fair, Class 42 seems to be where the entire
> concept of 'phishing' began -- trying to get account keys and
> usernames/passwords illegally.)  (Reference:
> http://www.uspto.gov/web/offices/tac/tmfaq.htm#Application018 , "What
> are the different classes of goods and services?")
>
> Why don't CAs have sub-CAs based on the class of the trademark that
> has been granted on the name?  (Trademarks often take over a year to
> process through the USPTO, but a 'no current trademark registered'
> sub-CA could be used for that.)  This would allow for determination of
> what class of business the name is certified for, and a UI enhancement
> that would thus prevent the GeoTrust attack of "Chase Ferries" being
> UI-same as "Chase Bank".
>
> It would seem to me that the class of threat that GeoTrust exposed was
> "right to use the name".


Oh, I see.  What you are essentially asking is
why don't they set up specialist CAs that do a
particular sector and leverage off the trademark
infrastructure and charge for that?

The reason for this is simply because the market
for CAs is inefficient and small.  Now, what follows
is based on econ, esp. Porter, so apologies in advance.

There is no free entry into the CA business.  The
barriers include "best practices" of how to run it,
browser requirements for WebTrusts, digsig laws.
These barriers both mitigate against competitors
of the current insiders, and also against the
discovery of more efficient services.  As the cost
of a WebTrust is put around the $100k mark, the
basic cost of a CA is looking around the $1m mark
which students of business will recognise as a
serious barrier.

Further, the barrier that was built was created on
a poor understanding.  It was constructed on paper
in advance of experience, and turned out to be
misaligned with market needs.  Unfortunately, a lot
of capital was raised on this design, and it was
invested into "best practices" sales, again in
advance of the experience needed.  (A metric for
this would be the Verisign stock market price over
the years.)

What this meant was that the market stalled, and
only those parts that were mandated ever happened.
Luckily (some would say) one essential part of the
PKI was constructed by Netscape which put it into
its browser, and this forced merchants to buy
Netscape's secure server product.  They needed
something to sell at the time, and this was one
nice possibility.

Other competitors
looking to take Netscape's market adopted this
complete solution without question, and within a
year or two the solution was spread across many
server suppliers and many browser suppliers.

Which caused a lock-in of the browser's use of SSL.
Merchants found that they had to purchase certs and
the market for certs became strongly correlated
with the retail ecommerce market (it tracked the
dotcom boom.)

But, most other users ignored it, which was what
all the capital was raised upon.  (These days you
will see some webmail users and all finance users
doing it, but even counted up, that makes for not
many certs.  See securityspace if you want to see
how few certs are sold every year.)

Which meant that the market for certs was locked
into a small mandated need, with no ability to
modify its approach.  Which means there is no
money to be made in a vertical, because the capital
cost outweighs the small number of certs.

If none of that makes sense, then think of it this
way:  The market is too small, and it is too small
because it is inefficient.

iang
_______________________________________________
mozilla-crypto mailing list
[email protected]
http://mail.mozilla.org/listinfo/mozilla-crypto

Reply via email to