/. [Intel Adds DRM to New Chips]

[Adam Back]

On Thu, Jun 02, 2005 at 11:05:30AM +0200, DiSToAGe wrote:
> I have read infos that say that audio and video drivers will be in the
> trusted chain. If your hardware system is used by an os (i.e. win) on
> which you can't create drivers, and only industry signed drivers can be
> used you can't bypass this by hacking drivers ...

Right.

> My though is the hardware drm can be reverse engineered ? If you use
> cert on your DRM you must put cert and private keys on your DRM chip ...

No the private key would be generated on the chip at manufacture, and
a signed certificate of it inserted by the manufacturer.

> So you can make a "soft drm" that use all the instructions of the
> reverse engineered hard drm, you but the reverse engineered private
> key, certs on your soft drm. 

It is feasible in the following way to make a soft drm.  

Step1. Get yourself a software controlled key signed by the hw
manufacturers.  Either:

1a. extract an already signed one out of the DRM hardware on your
machine by hardware hacking.

1b. find an insider at the manufacturing plant to sign a key actually
in the control of software;

1c. obtains the CA key used to do the signing (probably rather hard,
obviously they'll be trying to keep that one secure in tamper
resistant hardware with no key export function).

Step2. share the key, or setup a service to falsely authenticate
pure software DRM as hardware DRM with your key.

Now to stop you sharing this key directly or making a p2p DRM auth
server, they have to revoke the key.

I believe their revocation model is a bit weak from what I read of the
specs a while back.  They have a kind of challenge:

- to avoid criticism of privacy invasion, they have to make the thing
anonymous (or at least pseudonymous with lots of pseudonyms)

- however you can't blacklist a truly anonymous challenge-response.

(There was a protocol from Ernie Brickell with this kind of problem.)

Depending on what the final details are therefore their revocation
model might be weak.

> (so seems happy futur, something you buy and use but don't own ?)

Yes.  It is outrageous for the RIAA/MPAA and hardware companies to be
trying to foist this stuff on people.


The other way is to find a buffer overflow or such in one of these
privileged signed drivers and then you can inject code/or bypass DRM
restrictions in pure software.  They might at some point giving you
signed AND encrypted drivers so you can't even reverse-engineer them,
but I would say you have a right to know and control what is running
on your machine.

Another even more powerful buffer overflow would be one in the
supervisor / mini-OS that is hosting the Trusted Agents in ring -1.

Adam