Re: Fwd: Challenge: Confinement

[Michal Suchanek]

On 8/29/06, Christian Stüble <address@hidden> wrote:
> Hi,
>
> sorry for the late response, but I have to do some work in parallel :-)
> Therefore I may not be able to answer all questions. But I do my best.
>
> > In the context of personal data protection:
> > What kind of use do you have in mind?
> Depends. Of course a "not so efficient" approach would be to prevent someone
> fro copying my email address but display it on the screen. You can prevent
> that someone prints the address, but not that it is written down.
> Nevertheless, it would make it harder for 'parties' to give my email to
> someone else.

But you still cannot prove they didn't which was iirc the purpose of
the whole excerscise.

> Maybe more interesting applications I have in mind are:
> - Agents that store private information internally and use them only
> internally as an argument of an internal function. Some kind of
> object-oriented approach..
> - More pragmatic: Store a signature key inside that signs emails before
> sending it to you. You can define how often the signature can be used. Use
> the signature to identify non-spam.

And what does that do for you?
You can trivially recognize the emails that came from that system even
without TPM.

> > How do you enforce once-only
> > use?
> If you have a TPM (at least v1.2), it should be possible to prevent replay
> attacks. This functionality could also be used to enforce once-only use.
>
> > Once you get the data, you can print them, or write them down.
> > What kind of use guarantees no reuse?
> If (i) nobody has access to an application's internal state and (ii) the
> application decides not to print, it will not be printed. The first
> assumption is, of course, very important to enforce my personal privacy
> rules.

Unless the data never leaves your DRM system it _can_ be reused,
traded, and whatnot.
If it cannot leave the system (even displayed on screen) it is of very
limited use.

> >
> > If the administrator of the system cannot access the data how do you
> > make backups?
> The administrator may not be able to access the internal state of some
> applications. Nevertheless, it may be able to backup encrypted data. The
> challenging question is to backup information of type "replay attack
> protected". :-) We are currently thinking abou how to realize this in a
> multilaterally secure way.

The very purpose (or at least one of the purposes) of DRM/TPM
solutions is to prevent people to make backups (=copies) of the data.
It exists to enforce that the data (such as a movie) can be used only
on a specific machine, a specific number of times.

If you allowed backups that can be restored when the original system
is lost (which is one of the most important reasons for backups) you
would lose control over the use and dissemination of the data.

> > I do not see how DRM can be of much help if you want to use a system
> > that is controlled by a party that you do not trust.
> It is controlled in such a way that the remote party can define any security
> policy. Nevertheless, my "privacy-protecting agent" will only be executable
> if the security policy fulfills some of my requirements (e.g., not to access
> the state of my agent).

OK, while encryption allows you to store data on untrusted systems
(and while the data may be lost, it is unlikely to get disclosed or
modified) DRM allows you to execute jobs on untrusted systems (and
while they may never finish, they sould return correct results if they
do).

In return, you give content providers (and just about anybody) the
right to control how you use your computer. To access their
data/communicate with them you have to use certified OS, certified
application, in a certified way.

Of course, you have the choice to use the OS you like with the
applications you like in the way you like. But you will be able to
communicate and share data only with free software enthusiasts.

> > Sure encryption
> > can do something for you. DRM can do a little but not much. And you
> > still have to trust the provider DRM which I do not consider much
> > wiser than trusting the party controlling the system.
> What do you mean with DRM? What is the provider DRM? I am only talking about
> my provacy agent that is using TC-like technology to be able to negotiate
> a policy acceptable by my and the platform owner.

The DRM providers are in fact several.

There is the chip provider that certifies by a key that the chip is
indeed a valid TMP chip. There is nothing stopping him (or some of his
employers/contractors) to make invalid chips or give away keys you
could you to emulate a chip.

Sure it would harm their good name. But it turned out that providers
of SSL certificates did not mind their good name that much.

Then there is a party that certifies that the OS is genuine and
correct. The chip collects the OS signatures but you have to verify
them somewhere.

Then there is a party that certifies the application. Not all
applications would come with an OS so there will have to be some
separate application certificates.

Thanks

Michal