David Ross said:
> Stephen Davidson wrote [in part]:
> > 
> > Apple appears to have a slightly different policy, which is that if a CA is
> > accredited or regulated in their home country, then they should be included
> > in the OS.
> 
> I have a problem with this.  Some nations are rife with corruption
> and fraud.  For example, would you trust a CA certified by
> Nigeria?


David does raise the difficult issue in all this.  He wants
a certain quality standard imposed over well known unruly
peoples, such as Nigerians, which are the source of most 401
scams (indeed, the 401 scam is named after the section in
the Nigerian code, or some such.)

Yet, how are the Nigerians to deal with this?  Are they damned
for all time?  Or could they could set up their own CA and have
their government audit it and ensure its security, in order to
tackle the very frauds that bedevil the Nigerian reputation.

(As an anecdote, I'm involved with a conference in payments
and security, and, so far, by far the most registrants are
Nigerians ... Now, is this a bona fide effort to increase
payments security?  Or, is it an attempt to weasel a letter
of invite so they they can get a visa?  As the credit cards
these guys are providing aren't in their names, I'm betting
on the latter!)

Either way, a CA from Nigeria, authenticated by the goverment
there, could be bad.  Or good.  Tricky to say without some
more work.

Meanwhile, it seems that Microsoft has another solution -
shift the burden across to an independent audit standard.
Is that good?  Depends on the standard.  Depends on the
auditors.  Depends on Microsoft.  Depends on the money...

We've seen plenty of commentary, and to my mind, it all
seems to come down to ... it depends!

I don't think it's possible to say that Microsoft's method
will always be good, or Opera's or Netscape's.  Or that
Nigerians are all bad.  Or, any other group, either.



The audit process does have some bad points.

The independent audit will probably knock out CAs like
CACert.  It's just too darn expensive.  Then, there will
be CAs who can afford it, and they will be very happy to
make sure that it is hard to do, and expensive, simply
because that's one easy way to reduce competition.

Will it do its job?  Tricky to say.  On the one hand, we
have the supposed claim of high quality.  On the other hand,
we have the loss of faith in general audits over the last
few years.  OT3H, the expense of the audit process (and the
expense of the whole CA thing) does sit rather oddly with
the whole reason, foundation and spirit of the Internet.

One thing that audits *do* achieve (observably) is that
they reduce competition on qaulity.  That is, if everyone
who can play has to have an external audit, then there is
no incentive to do any more than the audit requires, and
only a sufficient incentive to do just enough to pass the
audit.  "We are audited by BlahBlahTrustBlah!" is the
beginning, middle, and the end of the security story.

This is the trap that the financial markets are in.  As
regulation is so stringent and so overwhelming, the notion
of competing on the quality of governance is totally lost
to the financial side of the markets.  I.e., everyone
does what is required and no more.  Partly because there
isn't any room to move from the regulations, and partly
because nobody rewards anyone for anything more than a
standard audit.

(Oh, and real security observations are punished.  If
one criticises an audited firm, one is rounded upon for
treating the regulators, the auditors, and the whole
darn system as if it was anything less than ordained
from on high.)



Some people have handed over their "trust" to auditors,
which is why they were shocked, simply shocked, when
Arthur Anderson failed.  US Congressmen were appalled,
simply appalled that their trust could be abused, so
Messrs Sarbanes & Oxley fixed it with more regulation,
more rules, more punishment, more audits, and ... less
security.

(David points out, that is finance, and this is process!
With respect, no, it's audits.  There is no difference
between what is audited.  The rot that is in the audits
is there for either.  You pays your money, and you gets
your audit.  If you pay enough money to your auditor,
you can get him to say what you want.   You don't, and
he'll say what he wants.)

Which is a bit of a criticism against audits.  In biz
speak, this is called *shifting the burden*.  That's
the name for when we've created some smoke-and-mirrors
solution that sounds good, but just moves the problems
from the left to the right.

Are the audits any good?  Who knows.  I say not, others
say yes.  I'd have more faith in a Nigerian CA if I knew
something about it, than some unknown cert with an unknown
CA.  If I don't know who the CA is, to me, it's a dead
issue.  That's because I've seen too many audits and/or
auditors, and/or a few CAs, that were jokes at best, or
frauds at worst.



Back to Mozilla and its CA policy.

How to get out of this mess?  Well, the current HTTPS
arrangement - the structure, the design, the architecture
- is broken.  In many ways, but one particular way, and
that is that *all* CAs are treated equal.  In this context,
the CA policy is doomed, unless this weakness of the HTTPS
CA implementation is taken into account.

"All CAs are equal" simply cannot work in a security 
framework.  It makes no business sense, as businesses
and people don't trust Nigerian government the same way
they trust Arthur Anderson.  And, it makes no security
sense to shift the burden from "untrusted but individual
merchants" to "group of CAs trusted by global fiat."

Hence, the suggestion (by many) that we need to stop
treating the CA as a commodity.  And start treating it
as a business:  With reputation, brand, and an ability
to suffer in the marketplace if they muck up.

Separate out the CAs.  If that is done - if the user gets
to see the CA when on the SSL-protected site - the users
can do the rest (with a lot of help from the market...).

If they can't, no policy by Mozilla, or Microsoft, or any
one else (DHS, pay attention!) is going to make it worth
the effort.  (Which is fundamentally why people call for
*lowered* standards, because they recognise that high
standards don't deliver, so let's save some money.)

Without being able to separate out CAs, it really doesn't
matter what policy is picked - one CA as good as the next
so it's a crap shoot.  All quality becomes the same.  Too
good for some, and too bad for others, so, the last thing
it is, is quality.

Without CA branding, CA policy is just a (nother) barrier.
Until CAs become known - in users' faces - and punishable
for delitos, then no amount of auditing will help any.
But, it might raise prices and shut out better players and
more secure solutions.

With CA branding, CA auditing becomes a competitive tool
for (some) CAs to compete with.  That might be a good thing,
and the market will decide.  But without branding, CA audits
have nothing.

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

Reply via email to