Re: Shifting around the risk burden, Re:

[Ed Gerck <egerck@xxxxxxx>]

All:

This is a reply to a private comment I received,
which may be useful here.

Cheers -- Ed Gerck

-------- Original Message --------
Subject: Re: Shifting around the risk burden, Re: Consensus Text for
Date: Mon, 13 Dec 1999 00:41:01 -0800
From: Ed Gerck <egerck@nma.com>

,,, wrote:

> I doubt the correctness of the quoted assertion from X.509:
>
> >A number of practical methods
> >are available for the user to hold his private key in a
> >manner that provides adequate security
>
> It depends on what "adequate" means, but there are too many published
> attacks on known systems to leave me at all comfortable with the practical
> methods I'm aware of.
>

Yes, that affirmation is weak -- and it was put there IMO in order to
try to preempt doubts. However, in discussing X.509/PKIX in order
to try to develop a better base for interoperation, which is my 
main objective, I have found it more useful to hold to one part of
the PKIX spec while verifying fallacies in another -- rather than trying
to show that both (or, more) parts are wrong.

What is at stake here is the Grandma scenario, where Denis insists that
X.509/PKIX can be used to deal with the content of what was signed,
as well as its intent, which are respectively second- and third-order removed
from the act of signing a message *digest* -- which is all that X.509/PKIX
deals with.  The fact that I am fully aware (assuming this much) that I am
signing a message digest, does not mean that I agree with the content
and also does not mean that I had the intent of declaring my definitive
agreement to it, irrebuttably.

So, first, I need to show that (above) -- then, the fact that not even
that modicum of trust is granted. I have to leave to a separate argument
the fact you point out That I cannot assume I was fully aware of signing
a message digest -- my key might have been stolen, for example, and I
might have never signed it.

Now, if I would question this first then people may say that I am being
"impractical" because everyone knows that we all use credit-cards on the
Internet and it mostly works, that business has risks anyway, etc.

So, the real problem needs to be tackled first IMO -- that X.509/PKIX is
not about validation of message *content* or *intent*, it is about
authentication of credentials as data.  The semiotic confusion is between
the message as data (syntax -- that carries values, which are always
uninterpreted) and the message as "code" (semantics -- that carries
instructions, which can only be functional after interpreted according
to a trusted context).  X.509/PKIX  deals with data, not with "code"
--  X.509 is moot on validation procedures for "code" aspects.  Thus,
it cannot be applied to draw conclusions on that which it outrightly 
does not represent.

Cheers,

Ed Gerck