3 ms·
While I don't agree with the FSF on even close to everything regarding trusted computing, I think for a fair discussion you'd have to at least steelman their ar
by lxgr 2y ago
While I don't agree with the FSF on even close to everything regarding trusted computing, I think for a fair discussion you'd have to at least steelman their arguments here:
I think it's fair to assume that in a world in which almost every device supports attestation and makes it available to any service provider by default, without giving users an informed choice to say no or even informing them at all, service providers are much more likely to provide access exclusively to attestation-capable clients.
That, in turn, has obvious negative consequences for users with devices not supporting attestation (whether out of ideological choice, because it's a low cost device and the manufacturer can't afford the required audits and security guarantees etc.): Sure, these users will always be able to just refuse to transact with any service provider requiring attestation.
But think that through: We're not only talking about Netflix here. At what availability rates of attestation will decision makers at financial institutions decide that x% is good enough and exclude everybody else from online banking? What about e-signing contracts for doing business online? What about e-government services?
I am at the same time excited about the new possibilities attestation offers to users (in that they will be able to do things digitally that just weren't economically feasible for service providers, since they often have to cover the risks of doing so) as I am very wary of the negative externalities of a world in which attestation is just a bit too easy and ubiquitous.
In other words, the ideal amount of general purpose attestation availability is probably high, but significantly below 100% (or, put differently, the ideal amount of friction is non-zero). Heterogeneity of attestation providers can probably help a bit, but I'm wary of the inherent centralizing forces due to the technical and economical pragmatics of trusted computing.
- fc417fc802 2y agoThe ideal amount of attestation on a general purpose computer which is owned by me is zero. Any nonzero amount implies that control of the device has not actually been turned over to me. It implies not only the slippery slope to which you refer but also things about back doors and opportunity for dystopian political regimes and much more. When it comes to financial or legal matters (and this includes online banking) a small dedicated hardware element for signing fingerprints is all that's ever been required. Anything more is an overreach.
- lxgr 2y ago> back doors and opportunity for dystopian political regimes No, this is a misunderstanding of what a TPM is. A TPM is a secure element inside your computer, similar to the chip running your credit and debit card. That's it. Without you using it (i.e. your OS or an application you installed asking it to do something), it's exactly as dangerous as a blank chip card in your house that you don't use and didn't open any account for. If you don't want anybody to talk to it, don't install applications or OSes on your computer that do things you don't want. You have full control over that! Not running software that's not acting in your own best interests is generally good practice anyway, TPM or no TPM. > [...] a small dedicated hardware element for signing fingerprints is all that's ever been required [...] You might be happy to hear that that's exactly what a TPM is, then!
- fc417fc802 2y agoI am fully aware of what a TPM is. I was speaking about trusted computing - ie the "general purpose attestation capability" that you referred to above. As you say, a TPM alone can't do much of anything and doesn't pose much of a threat. Of course expanding the acronym - Trusted Platform Module - is a bit of a giveaway. They were always fully intended to serve as the root of trust for much more nefarious things.
- lxgr 2y agoPeople keep saying that, yet the only thing I’ve ever seen TPMs used for is full disk encryption and user authentication. Conversely, DRM is alive and well on almost universally TPM-less devices. By the way, all of your comments in this thread end up dead – I had to vouch for them to be able to answer. Not sure what’s up with that.
- fc417fc802 2y ago> the only thing I’ve ever seen TPMs used for is full disk encryption and user authentication. Aren't all device attestation schemes underpinned by authenticated boot which itself is underpinned by a TPM? This is certainly the case for Android - AVB is implemented on top of secure boot on all the devices I've ever owned (and Play Integrity, if I had ever permitted it to run, on top of that). Do I have some misunderstanding about the stack? > Conversely, DRM is alive and well on almost universally TPM-less devices. You mean software DRM I assume? Because the only TPM free hardware backed DRM that comes to mind is GPU based encrypted streams where the GPU does the decoding and final compositing locally. And even then the TPM-equivalent exists, it just isn't accessible to the end user. SGX can be used to do various interesting things without attesting the state of the broader system, but none of the examples that immediately come to mind feel much like DRM to me. > comments in this thread end up dead Thanks for letting me know. I guess I should email them?