I think what you're saying was true in the past, but the game is
changing with modern hardware. There have been advances in CPU
features that make it possible to reduce the trust perimeter to just
the CPU and TPM. If I trust those two components, I can privately
compute on remote hardware, even if you have physical access and time.

If my computation involves sending video out to a device in the clear,
then yes, you'll get that video. So, this doesn't help with
traditional DRM. But it does help protect me when I run in
environments outside my control, e.g. in infrastructure-as-a-service.

Note, if an adversary can subvert the CPU itself, you lose. There are
also known flaws in TPM specs and implementations, although these have
a higher bar to exploit than other low-cost physical attacks.

Fortunately, there are some interesting CPU features in the pipeline
which may eliminate the dependency on TPMs completely. I can take it
off-list if you want to discuss further.

On Wed, Jul 31, 2013 at 7:32 PM, Danny O'Brien <[email protected]> wrote:
> Of course, such remote attestation/control works as well for
> privacy-preservation as it does for DRM -- not very. If you *have* the
> data, you can do whatever you want with it. My computer can attest all
> it want, but if I want that video or that cable, I'll get it. Indeed, I
> already *have it*. all you're doing is determining the process by which
> I'll obtain a reproducible copy. It's not even that quantifiable as a
> cost, because we're already assuming I have physical access and time
> enough.
--
Too many emails? Unsubscribe, change to digest, or change password by emailing 
moderator at [email protected] or changing your settings at 
https://mailman.stanford.edu/mailman/listinfo/liberationtech

Reply via email to