this is long and academic...delete now unless you are really interested...
>From: jueneman <jueneman%wotan(_at_)com(_dot_)gte>
>Subject: Re: Re: Soldiers, sailors, etc.
>Date: Wed, 12 Jan 94 17:13:43 EST
MOTIS is X.400 messaging in ISO terminology.
I do mean by key "all keying material identification and any associated
bits for use by ciphers acting to implement security mechanisms."
However, beyond that, the validation of a given statement is wholly
based upon the properties of the key distribution system, not
attributes (explicit or implied) which may accompany the keying
material. In validating one statement, I must not have to validate
another; only, assure myself that the key distribution service is
trusted.
The key distribution system must securely deliver an attributable
source which I may hold accountable to some standard, the quality of
which, together with the quality of my domain integrity-checking
procedures, may act to validate a statement purporting to offer a
messaging security service.
>
>Does "validate" equal "nonrepudiation" in your mind? Or are we talking at
>cross purposes.
I like the dicussion of non-repudiation given by Roe et al., though its
a bit academic: the set of beliefs held by the third-party acting as
arbitor will be used to decide a contended signature; one provides for
a non-repudiable signature when one shows that one intends to
substantiate the statement using the belief system of that party, when
known reasoning systems will be used to test that system's internal consistency
in the specific case.
We know that mere proof-of-origin might be asserted as: a given datum
was succesfully synchronously decrypted using a given key known to be
associated with mr X. Alternatively, there might be a TIS-PEM
certificate appended to a MIC, which one might check to learn that X is
affiliated to organization O, whose CA might be shown to a member of
the TIS PCA domain, which has stated procedures for mainitaining domain
integrity, and which is prepared to qualify the accuracy of the
affiliation statement.
However its yet another CA which would set itself up and assert: Our
beliefs and reasoning system for resolving disputes are the standard by
which the TIS PCA affiliation statements are themselves validated. If TIS PCA
certification policy matched that of CA then TIS certiifcates could be
used as case evidence. Alternatively, to be sure that the statement can be
checked by THAT CA, messaging users may explicitely obtain and exchange
CA's certificates, in order that there is no certainly no possible
clash of certification policy governing the applicability of evidence.
Now, it TIS set themselves up and assert not only the nature of its PCA
policy is assurance to level A, but also its belief system and
reasoning is suitable as a measure by which credible fourth-party
persons (judges!) could resolve an argument over the meaning of a
security service, then it could offer this property as a
consequence of using TIS PCA domain certificates. and this would be a
good reason for people to choose TIS for this quality of messaging
originator service.
Clearly, a business proposition in non-repudiation will only be viable
if there are reasons why subscribers would sign up to CA's beliefs and
values, based upon the recommendation of the judicial fourth-parties.
An easy route to solve this for general commerce and public is to
re-engage people trust in the things they have been brought up to
trust, automating where necesary in order to operate the
non-repudiation service. We all have trust in existing institutions and
structures.
Now in conclusion, I admit that the previous non-repudiation argument
actually contradicts my assertion that to verify one
statement one should not validate yet another. And then, I argued
that determining that one trusts the key distribution happens precisely
via that mechanism! However. this is becauses of the model
chosen to describe the system, based on the concept framework of
Roe et al. Steve Kent can model the same properties
using just the notion of a certificate being not
a statement of affiliataion, but a secure binding of...
The same relation between messaging service and key distribution holds
whichever model is used. And thats the point!