John S. Gage wrote:
> Simple (simplistic?) question.
>
> If the data are encrypted when they are collected, stored encrypted, and
> only decrypted by a known set of people with a known set of keys, then
> is it not true that:
>
> a) you are no longer reliant on the operating system or network security
> at all...
>
Yes, but this is far from current practice. My issues have been to 1) get people
to understand that current practice has limitations and where those are, and 2)
to get people to think about how the future of an all electronic record will
have to be different from a practical standpoint than today's practice.
>
> b) you have reduced security to a key management problem (stolen keys,
> borrowed keys, etc.)?
>
I think the key management issue is quite a bit larger than
you make it out. Let's say that I, a user, wish to access a
clinical record and that I have the access rights, the
privileges to do so. This data is encrypted with a key
known to only a few, but now I need access to that key. If
you associate my privleges with ability to decrypt the data,
hence access to key operations, I can move that data to any
system I wish, unencrypted.
So the act of encryption has accomplished what from a
privacy and confidentiality standpoint? It has become part
of an access control system, which is probably implemented
by OS security, so given a well run system, it is redundant
as a privacy or confidentiality mechanism.
The encryption has not guaranteed anything more than the
underlying privlege system can guarantee.
Now, if I change the model to having each user maintain
their own set of keys, I have the well known key management
problem which can be addressed by PKI systems. So now we
have each user maintaining their own private key part of the
PKI pair. How is that done? Well, take a look at the
practical systems in place today, and guess what, they use
some local form of authentication, probably using userid's
and passwords, implemented on insecure hardware and
operating systems. And for an enterprise, that's a mess, so
most enterprises resort to some form of network OS security
to secure the private keys. And we are right back to ground
zero again, only now we have layered on gobs of very
expensive and very complicated technology!
Fast forward to the future:
We have tamper-proof hardware low cost enough so that it is
everywhere. We have viable biometric's embedded into this
hardware along with a few well understood cryptographic
systems using that biometric information as a 'key' to
perform crypto operations and identity. Now we no longer
rely on userid/password systems at ground zero on untrusted
hardware/firmware/software. The whole promise of the
cryptographic techniques might become practical.
Fast rewind back to today: We have a long way to go...
Go out to your library or bookstore and read a copy of
Secrets and Lies by Bruce Schneier, it's not about math and
techological capabilities, it's about people and systems.
See: http://www.counterpane.com/schneier.html
and while there, subscribe to Crypto-Gram via e-mail, you
won't regret it.
