the repeated claim in this (and other threads) is that X9.59 eliminates
secrecy as requirement to guarentee the integrity of the financial
infrastructure (the requirement given the X9A10 working group for the X9.59
standard for all electronic retail payments).

the authentication requirement with x9.59 turns all retail payments into
authenticated transactions, eliminates any requirement for any
identification information and removes the characteristics of the account
number from being a shared-secret (i.e. the account number is currently has
the status of a shared-secret, since just by learning the account-number it
is possible to execute fraudulent transactions).

The business process issues are

1) is authentication or identification required (as per the original
subject line). this can be carefully considered. using the same basic
technology, it is fundamentally possible to implement either. it is not an
either/or possibility. the government could choose to reliably make your
digital signature equivalent to your identity. that is a government and
business process choice. however, it is not intrinsicly part of digital
signatures. It is possible to create a digital signature infrastructure and
protocols that are agnostic to whether the government has created an
infrastructure making digital signatures and identity equivalent. The issue
that the original subject line highlighted is that frequently the
technology of authentication and the implementation choice regarding
authentication vis-a-vis identification can be grossly confused and lead to
inappropriate implementations. This is one of the reasons that most
institutions backed away from the original x.509 identification certificate
paradigm, the inclusion of huge amounts of privacy information and
indescrimently spraying it all over the world was concluded to not be a
good thing.

2) certification is basically a business process that is independent of
whether it is deployed as a online process (with online transactions) or
manufactoring a stale, static certificate at some time in the past as part
of an offline paradigm. The certification body and business reliance can be
exactly the same in both cases. The difference is whether the relying-party
relies on a stale, static certificate or relies on a real-time, online
transaction.  It can be trivially shown that in almost all cases the
online, realtime responses is both higher quality and more valuable than
reliance on stale, static certificates. The absolutely worst case scenario
is that the high quality, realtime response is the same as stale, static
certificates. So one issue, is does the business justificy the cost of the
higher quality operation with an online, realtime response vis-a-vis the
possibly lower cost of the stale, static certificate information.
Certificates were original targeted in the early '80s when there was not
ubiquitous, inexpensive, easily available online infrastructure. As the
alternative to having no certification (having NOTHING), certificates were
better than NOTHING. However, the trade-off has significantly changed
during the technologies of the later 90s with the penetration of
inexpensive, ubiquitous, pervasive online infrastructure. This is
essentially the transition that the online financial infrastructure made in
the '70s.

In numerous past discussions, it turned out that many of the issues with a
global infrastructure didn't truely require identification, but it was
sufficient for the specific business processes that strong authentication
was sufficient and identification was mostly unrelated. For instance, it
isn't necessary for all parties even remotely involved in a financial
transaction to absolutely know who you are (aka identification) if strong
authentication is sufficient to establish that you are the entity entitled
to withdraw funds from a specific account (authentication). In fact, one of
the additional inducements for financial institutions to retrench to
relying-party-only certificates with absolutely no identification
information was an EU directive that said that retail electronic
transactions should be as anonymous as cash; specifically that at the point
of sale, the retail financial transaction needed to be agnostic with
respect to identity. That didn't preclude the government from creating some
identification procedure but it did direct that the actual financial
transaction was agnostic with respect to identify.

As per the original subject line, authentication is about determining is
the entity entitled to perform the operation .... without necessarily
requiring that the entity also be identified. It doesn't necessarily
preclude there being some process for establishing identification, the
authentication process can be agnostic with respect to identification.

There are large groups of people on both sides of the argument with regard
to having authentication processes that absolutely prevent any
identification vis-a-vis the authentication process doesn't preclude
establishing identification under specific guidelines. When given a
requirement to guarentee transaction integrity, strong authentication is
sufficient and KISS to meet the requirement.  Again that is the original
subject line; confusing authentication and identification.

However, the issue of offline, stale, static certificates as a
certification business process mechanism vis-a-vis online realtime
(account-based) certification is totally orthogonal to authentication
vis-a-vis identification issues.

The business issue for authorizing a transaction in conjunction with an
authentication mechanism does come into play a little. I've frequently
facetiously referred to an authentication retail store; you walk in
authenticate (or even possibly identify) yourself and walk in. The purpose
of the retail store is not to actual sell anything, it solely does
authentication (possibly identification) operations. This is opposed to
retail stores that are there to sell some item, authentication only comes
into play when performing a financial transaction (as part of paying for an
item to buy). There is nothing that is intrinsicly part of the
run-of-the-mill financial retail transaction that mandates identification
(there may be other aspects of some retail transactions that mandate
identification, but with regard to integrity of the financial transaction,
strong authentication is sufficient).

Now, each one of these environments tend to have a specific business
context which can be taken into account with regard to an online, realtime
certification. The certification authority can examine what is the context
of the business process and make the appropriate authentication with
respect to that context; aka rights regarding logging as a specific userid,
rights to doing financial transactions on a specific account, etc.  There
could be possibly hundreds of contexts that an online, realtime
certification authority could be agile enough to adapt to.

The issue with a stale, static certificates that are manufactored at some
time in the past with little or no fore-knowledge as to the business
contexts is quite a bit more problematic. There are basically three choices

1) relying-party specific certificates, one for each possible business
context, each one providing the specific authentication context (i.e. is
the entity entitled to perform that specific operation).

2) we periodically run across things jokingly referred to as GBD ... Great
Big Database in the sky ... the mother of all databases that contains all
possible pieces of information. This is a certificate that contains the
necessary authentication information for every possible business context
(is the entity entitled to do the specific operation in every individual
business context).

3) switch from an authentication (whether or not the entity is entitled to
perform the operation) paradigm to an identification paradigm. The
assumption is that all validly identified individuals are entitled to do
everything, possibly a log is made of their activities and there is some
settlement made by the government at the end of the month (or at some other
interval). It then isn't necessary to provide all possible authentications
for every possible business context. Everybody, once identified, are
automatically entitled. The stale, static certificate is then reduced from
having to serve any authentication purpose and is just relegated to only
having to perform an identification role.

Trying to establish the value of a PKI stale, static certificate-based
operation gets into this muddle. Huge numbers of certificates, one for each
context means that the individual value of any specific certificate is
significantly depreciated.  A single stale, static humongous GBC (great big
certificate in the skly) gets to become unwieldily having all necessary
information for authentication in all possible business contexts (aka
whether the entity is entitled to perform the operation). The GBC is also
likely to represent more in business costs than can ever be recovered in
certificate pricing. So it quickly comes down to that any sort of major PKI
infrastructure is pretty much stuck with an identification paradigm. It
isn't so much a business characteristic of certification (which i've
repeatedly asserted is totally independent of whether it operates as
offline, stale, static certificate, or a realtime, online, certificateless
operation). Any major effort at PKI with stale, static certificates pretty
much concludes in the identification model (aka #3 above). It is basically
what the x.509 identification certificate scenario arrived at in the '80s
and the early '90s (effectively because of the considerations outlined
above). It is also what was rejected in the mid to late '90s because of the
serious privacy issues involved in making a ubiquitous transition from
predominately authentication paradigms to a identification only paradigm.

The identification only scenario then still faces  the serious business
quandary of whether an entity is actually entitled to perform the specific
operations in the specific context (independent of identification). GLB,
HIPPA and other privacy-related legislation is also turning identification
into more and more of a business expense.

Finally, in normal business, there is a relationship between the provider
of something and the purchaser of something. The typical PKI business model
is almost a total violation of any existing infrastructure. In all of the
online models, the relying party directly pays the certification agency for
the online, realtime certification response. There is a valid, direct
business relationship between the relying party and the online
certification agency. Tbis gets to be really horrible in the traditional,
static, stale certificate scenario. The business relationship is between
the PKI and the public key owner that purchases the certificate. There is
no exchange of value between the certification authority and the
relying-party which would be the basis for a normal accepted business
process. In the normal accepted business model today involving a public key
owner purchased certificate, there is no exchange of value between the
relying party and the certification body, and therefor there would normally
be absolutely no recourse for the relying party (based on anything the
certification body did or didn't do).

So the traditional PKI paradigm with stale, static certificates left over
from the offline paradigm is faced with:

1) significant transition throughout the world from an offline environment
to a nearly ubiquitous online environment

2) the whole rejection of x.509 identity oriented stale, static
certificates in the mid 90s because of serious privacy concerns.

3) difficult to create any value proposition for a stale, static
certificate since there is no business relationship between the relying
party and the certification body (the certificate isn't for the benefit of
the certificate owner, it is for the benefit of the relying party ....
however the relying party that would benefit,  isn't paying and therefor
under normal business relationships has no recourse to whether the
certified information is valid or not).

So the next muddle is since that it really is difficult to establish a
viable business model .... turn it into a governement provided
identification service. In some sense this isn't an authentication
requirement, which can be handled perfectly validly with an online paradigm
and suffers none of the shortcomings of a stale, static certificate
paradigm.

In some sense, it is a extension of the line of thought that (stale,
static) "Certificates Are The Answer" and the government provided
identification service might appear to be the only viable way to create a
role for certificates. 1) Governments can mandate things and totally ignore
issues like the world transitioning to an ubiquitous online environment. 2)
Governments can madate things and totally rewrite what are acceptable
privacy guidelines and what are not. 3) Governments can mandate things and
totally dictate what relying parties will accept.

So lets say we have a government mandated  provided identification service
based on stale, static certificates. As outlined in the previous post:
http://www.garlic.com/~lynn/aepay11.htm#72 Account Numbers. Was: Confusing
Authentication and Identification? (addenda)

we have the case of doing ISP login using digital signatures. With the
stale, static certificate model, the router on receving the triple (login
transaction, digital signature, certificate) should have all the necessary
and sufficient information to establish the login operation. The
certificate gets validated as a valid, government identification
certificate and the public key from the certificate validates the digital
sginature; voila the person is connected w/o needing to find out if the
entity is entitiled or not (authentication). There is absolutely no
requirement for authentication since the government identification
certificate is sufficient for establish the right for service.

So we are led into this muddle that now has both 1) giveronment
identification stale, static certificates, and a government identification
stale, 2) static certificate is sufficient for authorizing all operations
w/o requiring any additional effort.

Now, unless we have created a totally government supplied world for all
possible services; there typically are business entities and they establish
some sort of business operation for services or products. Today these
business relationships are represented by account records. People having
contracted for ISP internet service have a RADIUS account record. The
logical account record (which may be partitioned into several physical
records) typically contains the T&Cs of the business relationship as well
as the necessary information to authenticate the entity that is entitiled
to use the contracted for service. In the traditional business
authentication-based scenario, the entity creates a login transaction with
the appropriate authentication information (which I assert can be a digital
signature) and the double (aka transaction plus digital signature; not the
triple as in the stale, static certificate based scenario) is transmitted.
The receiving router gets the username (or other) handle
and retrieves the corresponding RADIUS record. The transaction is
authenticated with the information in the RADIUS record and then the rest
of the information from the RADIUS record indicates the entitled to
service. Lack of the appropriate RADIUS record or lack of authentication is
an endication that the requesting entity has not contracted for the
requested service.

In the offline, stale, static identification certificate scenario, it is
sufficient to identify the requesting entity, it isn't actually necessary
to establishing that the requesting entity is entitled to the requested
service (or is entitled to the explicit or implicit permissions).

so back to the original subject line about confusing authentication and
identification:
http://www.garlic.com/~lynn/aepay11.htm#66 Confusing Authentication and
Identiification?
http://www.garlic.com/~lynn/aepay11.htm#67 Confusing Authentication and
Identiification?
http://www.garlic.com/~lynn/aepay11.htm#68 Confusing Authentication and
Identiification?
http://www.garlic.com/~lynn/aepay11.htm#69 Confusing Authentication and
Identiification?
http://www.garlic.com/~lynn/aepay11.htm#70 Confusing Authentication and
Identiification? (addenda)
http://www.garlic.com/~lynn/aepay11.htm#71 Account Numbers.  Was: Confusing
Authentication and Identiification? (addenda)

--
Internet trivia, 20th anv: http://www.garlic.com/~lynn/rfcietff.htm

[EMAIL PROTECTED] on 6/26/2003 12:31 pm wrote:

In the purest sense, an individual is a discrete entity
and why should their identifier be a topic of discussion?
In every advanced country we already have a name and
a government-maintained UID.

The financial industry (and others) needs to end their
reliance on secrecy of the UID and associated information
per se.

I am uneasy with Lynn's criticism of "stale" digital credentials,
since, how else is the individual ever to achieve the
capability to identify ourselves over networks?

At times I feel the internet may be truly, fundamentally useless
until there is a global and publicly accessible registry of
individuals and their public keys.  As parties are not strongly
identified then, communication about things of consequence
is proving to be impractical.   Everywhere I look, I see
large efforts and investments at the server side, to garner
some profit for (increasingly larger) central hosts.   I see
nothing intended to give the economic benefits of the global
internet to individuals or endpoints on the internet.

With great reluctance I am beginning to agree with the
European governments' notions of providing a PKI for
their citizens.  In the US, I believe it would be a good thing
if the federal government maintained a network interface
where anybody could submit a SSN, and receive in
return, the name, address, and public key of the individual.

Todd



Reply via email to