Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The GPU, not the TPM, is the root of hardware DRM

mjg59.dreamwidth.org

Post reply on HN
Print view View original

401–410 of 493 posts

Re: The GPU, not the TPM, is the root of hardware DRM

#401

Post by userbinator »

mixmastamyk wrote: post #116
mjg59 wrote: post #107

Earlier quoted context omitted.

I'm extremely familiar with the capabilities of TPMs (I've worked on deploying remote attestation services at multiple companies), but here's the thing - streaming vendors don't use TPM-based remote attestation. None of them. It doesn't happen. Could it happen? Yes, but it would buy almost nothing - remote attestation is something that's viable in enterprise environments where you can bind TPM identity to inventory e…

Why is the grandparent dead? Makes an interesting point and am glad I read it and the replies. Please vouch.

Look at the other flagged comments and come to your own conclusion.

Re: The GPU, not the TPM, is the root of hardware DRM

#402

Post by lxgr »

forty wrote: post #400
mjg59 wrote: post #391

Earlier quoted context omitted.

It could , but why? They've come up with a solution that avoids having to place any trust in the OS at all, so why introduce additional complexity and fragility?

I don't know, maybe because not everyone has the right GPU with DRM, while the TPM is conveniently mandatory? :)

Mandatory where? Most of the devices people are streaming video to these days aren't even PCs!

Macs haven't had TPMs for a while now (I think Apple never really used it and dropped it even before the Apple Silicon switch), but of course they have their proprietary equivalent.

Re: The GPU, not the TPM, is the root of hardware DRM

#403

Post by mjg59 »

forty wrote: post #400
mjg59 wrote: post #391

Earlier quoted context omitted.

It could , but why? They've come up with a solution that avoids having to place any trust in the OS at all, so why introduce additional complexity and fragility?

I don't know, maybe because not everyone has the right GPU with DRM, while the TPM is conveniently mandatory? :)

If they don't have a GPU that implements this then the decrypted material would be available to the OS, which is precisely what the streaming media platforms want to avoid.

Re: The GPU, not the TPM, is the root of hardware DRM

#404

Post by mjg59 »

cryptonector wrote: post #393
mjg59 wrote: post #358

Earlier quoted context omitted.

> What do you mean "too new"? Like, you replaced your TPM? That's a thing on servers, but not laptops. I buy a GPU in 2025. I buy a new motherboard in 2026 and plug the GPU into it. How does the GPU learn about the new EK CA? These are devices that can be moved between systems, you can't delegate this to the platform vendor or TOFU, the GPU would need to generate independent trust in the TPM.

One way I might handle this would be to have a TPM on the GPU itself. Then you can move the GPU about all you like, and it will work. The GPU would have to implement an API and protocol that allows the DRM site to do attestation via software running on the CPU, but that seems doable. The other way would be accept that the GPU that the content is to be played on might not be the same as the device on which the TPM exi…

Why do any of this rather than just have the GPU prove its identity to the streaming platform? You're adding a lot of complexity for no obvious gain.

Re: The GPU, not the TPM, is the root of hardware DRM

#405

Post by wkat4242 »

hathawsh wrote: post #306
hiatus wrote: post #254

Earlier quoted context omitted.

Have you been to an industry conference? So many laptops are covered in stickers, good luck recreating that.

I don't mean to disagree, but I think it's worth pointing out that with today's tech, it wouldn't be difficult for an attacker to also scan the stickers and print them out on sticker paper using a color printer, all in minutes. And the technology for doing that is only getting better. Just a thought.

You would sooo notice. Most of my laptop stickers are half on top of each other and really worn :)

Re: The GPU, not the TPM, is the root of hardware DRM

#406

Post by sidewndr46 »

osy wrote: post #26

> I'm going to be honest here and say that I don't know what Microsoft's actual motivation for requiring a TPM in Windows 11 is. It is quite obvious: to force people to buy a new PC. TPM provides no added security value for the vast majority of users[1] but it is a convenient hardware that has only started to become standard (fTPM) in PCs built in the last ~8 years so it provides an excuse for Microsoft to declare co…

I'm relatively certain one of the motivations is to avoid people using Windows on virtualization without their consent.

Re: The GPU, not the TPM, is the root of hardware DRM

#407

Post by lxgr »

no_time wrote: post #377
mike_hearn wrote: post #364

Earlier quoted context omitted.

This is an FSF level understanding. Android devices are fully open and you can reflash them to whatever OS you want. Some remote servers won't give you service if you do that, but nothing is locking you out of your device . As Android dominates the global market, you already live in that world where most devices are open.

>Some remote servers won't give you service if you do that This is exactly my problem. Before ideas like this surfaced, the demarcation line between who controls what was purely based on ownership. The machine that I own acts only on my behalf and in my best interests, the server that you own does so for you (or atleast for PCs this has always been the case) TPMs, attested bootchains and whatnot trample on this whole…

Fully agreed on attested bootchains. General-purpose level OS-wide attestation is indeed a blight on open computing: It's ineffective because it implies a gigantic trusted code base (what are the odds that the entire Windows kernel is completely free of vulnerabilities?), and conversely it does tie you to somebody else's more or less arbitrary kernel build.

Almost complete disagree on TPMs. A better comparison than a spy would probably be a consulate (ok, maybe an idealized one, located underground in a Faraday cage): Their staff doesn't get to spy on you, but if you ever do want to do business with companies in that country and need some letters notarized/certified, walking into their consulate in your capital sure beats sending trustworthy couriers around the world every single time.

To torture that analogy some more: Sure, the guest country could try to extend the consulate into a spy base if you're not careful, and some suspicion is very well warranted, but that possibility is not intrinsic to its function, only to its implementation.

Re: The GPU, not the TPM, is the root of hardware DRM

#408

Post by lxgr »

mike_hearn wrote: post #383
no_time wrote: post #377

Earlier quoted context omitted.

>Some remote servers won't give you service if you do that This is exactly my problem. Before ideas like this surfaced, the demarcation line between who controls what was purely based on ownership. The machine that I own acts only on my behalf and in my best interests, the server that you own does so for you (or atleast for PCs this has always been the case) TPMs, attested bootchains and whatnot trample on this whole…

It's not just you but what people who hate remote attestation tend to forget is that it's a sword that cuts in both directions. Servers can remotely attest to you, not just the other way around. Signal is an example of an app that demands a remote attestation from the server before uploading your sensitive data. Attestation is just a tool. It can be used for all kinds of things and doesn't privilege one side or anoth…

> The average app developer doesn't truly care what device you use, they just want to cut out abuse and fraud, which are real problems that do require effective solutions.

I don't doubt that. But the price of attestation, if it's not properly isolated from the hosting OS (like Microsoft's completely unrealistic attempts of bringing the whole OS into the trusted computing base, kernel and applications and all), would be a homogeneity of computing I don't think is necessarily worth the benefits.

The good news is that such proper isolation is not only possible but even desirable (it keeps the trusted computing base small), and if done well could actually replace annoying half-measures such as "root detection": Who cares if my phone is rooted, as long as my bank's secure transaction confirmation application is running in a trusted, isolated enclave, for example?

Re: The GPU, not the TPM, is the root of hardware DRM

#409

Post by kuschku »

mike_hearn wrote: post #367
kuschku wrote: post #321

Earlier quoted context omitted.

https://aliexpress.com/item/1005003020587234.html

That device makes no mention of stripping HDCP and I can't find any evidence it does that on the maker's website. What makes you think it will strip HDCP?

> I can't find any evidence it does that on the maker's website.

I believe that's intentional (it would be illegal to import if it was advertised).

> What makes you think it will strip HDCP?

I've got 6 in use for lecture/talk recording without worrying about HDCP. Especially useful for presenters casting to chromecast or presenters using macbooks with DRM software (as blackmagic SDI converters don't support HDCP at all)

Re: The GPU, not the TPM, is the root of hardware DRM

#410

Post by fc417fc802 »

matheusmoreira wrote: post #380
ignoramous wrote: post #165

Earlier quoted context omitted.

> The author is correct in that media DRM is tied to GPU vendors on the field right now ... hardware backed DRM can be so much more invasive I expect mjg59 to know what they're talking about but like you say, I wonder the same thing about the strength of (what you call) Media DRM v Hardware-backed DRM. GPU vendors have quietly deployed [hardware-based DRM] ... [which] works just fine on [boards] that [don't] have a T…

The GPU is a completely separate computer running proprietary software. "Operating systems" do not operate anything anymore. They are just some user app, to be sandboxed very far away from the real action. https://youtu.be/36myc8wQhLo Stallman warned everyone. Virtually nobody listened.

> "Operating systems" do not operate anything anymore.

Not entirely fair. There is still a kernel and a privileged userspace layer. That hasn't changed. The OS implements a common API that abstracts over ISAs and other finnicky hardware details that are under constant short term churn.

It's just that peripherals themselves have become so incredibly complex that many of them now require their own embedded systems in order to operate. The hardware was always a black box it's just that now it contains an entire embedded OS.

Post reply on HN