Ram A Moskovitz <[EMAIL PROTECTED]> writes:
> I wonder why Mastercard/VISA didn't and don't choose to run
> everything themselves in house.

Mastercard/VISA are brand associations. also when they started, there
were possibly 30k banks issuing and acquiring credit cards ... each
with their own processing.

basically there are two variables for four possible conditions:

* physical and offline
* physical and online
* electronic and offline
* electronic and online

when things first started, the operation was physical (plastic card)
and offline (paper processing). there were invalidation booklets that
were mailed to all merchants monthly containing all invalid/revoked/etc
account numbers.

To some extent the CRLs defined for PKIs were modeled after the
physical and offline paradigm invalidation booklets from the 60s. The
CRLs also somewhat differentiate a PKI operation from a purely
*certificate manufactoring* infrastructure ... a term we coined in
the mid-90s when we started working with this small client/server
startup that wanted to do payments
http://www.garlic.com/~lynn/aadsm5.htm#asrn2
http://www.garlic.com/~lynn/aadsm5.htm#asrn3

The problem with the 50s & 60s paradigm was the inability for the CRL
paradigm to scale-up ... the numbers of invalid accounts were
increasing dramaticly (and therefor the aggregate risk from invalid
accounts) and the number of merchants were increasing dramatically.
The result was that you had to send out much larger booklets to a much
larger number of merchants on a much more frequent interval (i.e. risk
is in part proportional to the distribution interval).  Pretty soon
you were looking at dristributing booklets with thousands of pages to
millions of merchants every couple of hrs.

Now the niche that the Certificatin Authorities, PKIs, and digital
certificaes fit into is the "electronic and offline" paradigm. The
payment industry bypassed that archaic step and went directly to
electronic and online; a magnetic stip was put on the same plastic
card and infrastructure was deployed to perform real-time, online
transactions. Not only did they get real-time authentication and
authorization but they were also able to get operations dependent on
data aggregation (like credit limit, patterns of useage that indicate
fraud patterns). The PKI model with a stale, static digital
certificate would have had them imprinting the physical card ... with
the card's credit limit ... but the stale, static digital certificate
PKI model has no mechanism for indicating outstanding charges and/or
whether the aggregate of past transactions exceed the credit limit.

Somewhat after we had worked on this thing that has since come to be
called electronic commerce ... there were a number of people
suggesting that credit card transactions could be moved into the
modern world by having card owners digitally sign the transaction and
attached a digital certificate. Our observation was that the digital
certificate attachment would have actually regressed the payment
transactions by 20-30 years to the 50s ... as opposed to moving it
into the modern environment.

The x.509 identity certification business of the early 90s was
starting to wonder what information a business or relying party might
actually need about the person ... not being able to accurately
predict what information might be needed, there was some direction to
grossly overload identity certificates with enormous amounts of
personal information ... in hopes that a relying party or business
might find some information in the certificate of some use.

By the mid-90s, some number of institutions were starting to realize
that x.509 identity certificates, grossly overloaded with enormous
amount of personal information represented significant privacy and
liability issues. As a result, you started to see some appearance
of relying-party-only certificates
http://www.garlic.com/`lynn/subpubkey.html#rpo

which basically only contained some sort of repository lookup value (a
repository that contained all the real information) and a public key.
However, it is trivial to show that such an implementation both
violates the original design point justifying PKIs and digital
certificates but that they were also redundant and superfluous.

Now with regard to the association brands. As the transition of the
payment card industry to a realtime, online processing model (as
opposed to the archaic offline processing model represented by PKIs,
certification authorities and digital certificates), an emerging
requireming was that all of the independent bank processing operations
needed to be interconnected. So one of the things that you started
seeing appearing was the assocations deploying in the 70s and 80s was
a value added network (in general, value added networks saw a big
increase in the 80s ... but have since somewhat died off, having been
obsoluted by the internet). The band associations have maintained
their value added network infrastructure ... even in the face of the
internet having generally obsoleted most other value added network
operations.

The current processing outsourcing to is not by the brand associations
... which still operate their value added networks to tie together the
multitude of bank datacenters (getting realtime transactions from the
merchant bank to the consumer bank and back) ... but by the banks
... where they have outsourced their whole processing operation. In
this situation, it is different from outsourcing certification and
trust ...  separated from the actual transaction processing (which
simple security theory indicates can lead to fraud) ... but all
aspects of the online, realtime processing has been outsourced.

So looking at a digital certificate related scenario ... the
ssl certificates
http://www.garlic.com/~lynn/subpubkey.html#sslcert

this small client/server startup that wanted to do payment
transactions (which is now frequently referred to as electronic
commerce) had this stuff called SSL that was to be used in the payment
processing. So as part of the deployment we had to go around and audit
the major institutions called certification authorities (this is when
we coined the term *certicate manufactoring* to distinguish the
operations from the thing that is commonly called PKI in the
literature).

so the justification for SSL certificates was because of percieved
integrity weaknesses in the domain name infraustructure and the
possibility that a client might be talking to a webserver different
than the webserver they thought they were talking to.

So a merchant applies for a SSL digital certificate for their
webserver internet name. Then in the SSL protocol ... the browser
validates the webservers SSL digital certificate and then compares the
webserver name that the client typed in against the name in the
digital certificate.

It turned out there are a couple problems. 

A lot of merchants found out that using SSL cut their webserver
performance by 80-90percent.  As a result they started only using SSL
for the payment/checkout phase of shopping. The problem is that the
browser no longer checks what the client typed in against a SSL
certificate. When the client gets around to payment/checkout, they
click on a button ... that does the URL stuff automatically. Now if
the client had been visiting a fraudulent/incorrect website site
(which SSL is designed to prevent), such a fraudulent website is
likely to have their payment/checkout button specify a URL for a
website name for which they have a valid certificate (defeating the
original purpose for having SSL).

Another issue is that most certification authorities aren't actually
the authoritative agency for the information being certified (which,
in turn, the digital certificate represents that certification
business process). In the SSL scanario, the SSL certification
authorites have to check with the authoritative agency for domain name
ownership. An issue is that the authoritative agency for domain name
ownership is the domain name infrastructure ... which has the original
integrity issues justifying the use of SSL.

So somewhat backed by the SSL certification authority industry, there
has been a proposal that domain name owners register a public key for
their domain name ... and all future correspondance is digitally
signed by the corresponding private key. The domain name
infrastructure then can retrieve the onfile public key to verify the
digital signature (as countermeasure against various fraudulent
manipulation of the domain name infrastructure) ... not this is a
certificateless operation:
http://www.garlic.com/~lynn/subpubkey.html#certless

There is also some additional benefit to the SSL certification
authority industry; they can replace the time-consumer, expensive and
error prone process of acquiring identification information from the
SSL certificate applicant and attempting to match it against the
identification information on file in the domain name
infrastructure. Instead they can have a much simpler, reliable and
less expensive authentication process ... where the SSL certificate
applicant digitally signs their application and the SSL certification
authority retrieves the onfile public key from the domain name
infrastructure and validates the digital signature:
http://www.garlic.com/~lynn/2005o.html#40 Certificate AUthority of a secured 
P2P network

Unfortunately this represents something of a catch22 for the SSL
certification authority industry

1) increasing the integrity of the domain name infrastructure lessens
the original justification for having SSL digital certificates

2) if the SSL certification authority industry can base their whole
trust root on using onfile public keys for digital signature
verification (totally certificateless), it is possible that the rest
of the world could start also doing real-time, online retrieval of
onfile, certificateless public key operation.

One could even imagine a highly optimized, certificateless SSL
protocol ... where the response to the hostname not only carried the
ip-address but also piggybacked any available public key and any other
applicable SSL protocol information (in the same transaction).

Then in the initial client session setup with the webserver ... they
can include the (public key) encrypted symmetric session key and the
actual request. The webserver then can both do session setup and
process whatever request in single operation.

Here is a scenario where improved integrity can bring any outside
certification operation intended for an offline environment back in
house ... drastically simplifying all aspects of the operation, moving
to an online, real-time paradigm, and improving overall integrity.

-- 
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/
_______________________________________________
mozilla-crypto mailing list
[email protected]
http://mail.mozilla.org/listinfo/mozilla-crypto

Reply via email to