Comment by naikrovek | Hacker News Reader

Comment by naikrovek | Hacker News Reader

Parent Full thread

> it was designed to provide security against the owner

no. no, no, no, no, no. holy crap, no.

no.

it was designed to prevent unauthorized access to hardware.

corporations own computers, too. and you know what is on a lot of the computers they own? trade secrets. who wants those trade secrets? competitors, copycats, and so on. what is something that such motivated actors have a large amount of? money. money to hire thieves. money to buy expensive equipment. people to attempt to circumvent security measures.

Pluton is designed to address situations like this.

no one cares about John and Jane computer owner. Pluton will not be used against John or Jane unless John or Jane allow it, in some unforeseen situation I can't imagine.

Pluton is a very good weapon for corporate IT admins trying to secure their devices.

"BUT TPMs ALREADY DO THAT"

not unless you require a PIN on boot, and PINs on boot are a pain in the ass. no one likes them, no one wants them. they often have huge complexity requirements and can be easily forgotten. PINs suck. they suck so much that some unfortunate companies decide against the PIN requirement and implement things like BitLocker using the TPM alone, and that is extremely insecure to anyone with physical access to a device because TPM traffic can be captured.

"that seems like something that is very difficult and will never happen."

it's very easy: https://news.ycombinator.com/item?id=29258879

Pluton was designed for the same thing that the Intel Management Engine and other related technologies were designed for: corporate IT administrators.

In other cases where enterprises wanted something that consumers cared little about (ECC RAM comes to mind, or the artificial limit on number of concurrent connections in non-server Windows editions), manufacturers were happy to cripple their product for consumers, so they could charge a premium to less price-sensitive enterprises.

But when it comes to Pluton, or the AMD and Intel management engines, they're making sure you can't buy a product without them at any price.

> it was designed to prevent unauthorized access to hardware.

Yes and it's all a big coincidence there's no owner override, so that only the manufacturer decides what is "authorized".

I'm talking about PCs.

the administrator gets to decide what is turned on and how it is used. the administrator has back doors in, that are more involved and more secure than the front door. the private keys don't go to the manufacturer, but are created by the owner, etc.

there is always owner override in the case of the PC.

consoles are locked down because we have demonstrated to console manufacturers again and again, for 40 years, that we will pirate games for the console if it is at all possible. we have left zero room for interpretation with this message. we WILL steal if there is any way to steal. so in the case of consoles, yes. manufacturer only.

> there is always owner override in the case of the PC.

Oh, the owner can tell Pluton to fake attestation results, so that, for example, it would report to a program that it's being executed in Genuine™ Microsoft Windows® 11, even if the program is actually being ran in a sandbox on linux?

Because if not, then there is no owner override, and it's a scheme to foist hardware-backed DRM onto PCs.

BitLocker type solutions don't rely on remote attestation. RA has never really taken off in any context.

Regardless, arguments against remote attestation are always very tricky. You're acting like it's morally clear cut but it's not.

RA is designed to stop you lying about what you're running to other people. It doesn't stop you running things, but it may mean others choose not to trade or engage with you based on what you're doing with the computer. So this is a direct translation to standard libertarian philosophy. You can choose your own terms but you can't force your preferred terms on others - you have to negotiate a contract that's mutually acceptable and then that contract needs to be enforced. RA is a kind of technologically enforced contract, instead of relying exclusively on the legal system.

That's why arguments that RA is evil always skip a few steps. The argument goes:

- I want to run X (typically some Linux distro)

- Some service Y won't trade with me unless X will do Z (typically enforcing some sort of agreement e.g. anti-cheat or license).

- X chooses not to do Z but I want Y to service me anyway. So I want to lie to them and feel entitled to being able to do that without being caught. This is moral because Linux.

Often with a helping of "I'm afraid that actually service Y might be most services".

The concerns here are valid - it would suck to no longer be able to use Linux with various remote services - but if you look at how the TPM and related tech are designed there's no actual technical reason you couldn't make Linux satisfy remote attestations. It's simply that Linux users would usually refuse to do so on philosophical grounds. Well, that's fine, but from a political and philosophical perspective it's reasonable for your counterparty to want some assurance that you'll actually implement the agreement you're claiming you will.

> So this is a direct translation to standard libertarian philosophy. You can choose your own terms but you can't force your preferred terms on others - you have to negotiate a contract that's mutually acceptable and then that contract needs to be enforced. RA is a kind of technologically enforced contract [..] Some service Y won't trade with me unless X will do Z (typically enforcing some sort of agreement e.g. anti-cheat or license).

What can I say but thank you - you've distilled the problems with the TPM much better than I have. What you describe is exactly what I fear. Look at how these contracts are "negotiated" today - the consumer either accepts, or goes to live in the woods ("I'm afraid that actually service Y might be most services".)

And the terms are one-sided at best, usually curtailing a bunch of important rights (mandatory arbitration, bans on publishing benchmarks, anti-reverse-engineering or indeed any kind of in-depth examination of functionality, such as when Facebook barred researches from gathering data from volunteers on what ads they were showing, bans on scraping publicly-accessible data - good luck comparing prices,.. you get the picture)

So what kind of contracts do you think we'll negotiate, without even the thin pretense of a fair legal system protecting us, and when "our" machines enforce their unbreakable chains?

If I'm being charitable, I'd put the odds that the overall effect of this expansion of contract law will be to our benefit as "slim".

>when it comes to Pluton, or the AMD and Intel management engines, they're making sure you can't buy a product without them at any price.

Not only that but under the guise of "recycling", many 18-wheelers of perfectly functional PCs are being destroyed rather than re-purposed and these are primarily the ones powerful enough to run the latest OS's.

Anything decently powerful lacking the encumbrance of the hardware ME/TPM generation is the prime target of the agressive destructive effort.

You picked a bunch of conditions you don't happen to like and then projected this onto the technology along with a giant helping of victimhood, but again, RA is morally neutral. It's like encryption, it can be used for things you may agree or disagree with but the tech is just a tool. In this case it's just a way to enforce agreements.

So here are some agreements that maybe you'd actually like to enforce on other people, using the exact same technology.

1. Cloud privacy. RA and related tech let's you request a remote attestation from the cloud provider that proves your VM memory is encrypted such that the hardware provider can't read it. This means you can compute in the cloud and not be spied on, not even if the legal system in the remote jurisdiction suddenly demands it.

You might think cloud providers would never offer this but you'd be dead wrong. Both Azure and GCP offer RA in which you are the one checking them and they offer it today.

2. Blocking spamming without needing captchas or modern JavaScript based equivalents. In turn that means you can now much more easily self host email and retain control and privacy currently sacrificed to big email providers.

3. Copyleft enforcement. Yes really. It'd require toolchain integration but it's possible.

4. Eliminating backdoors inserted by cloud CI systems. They prove what compilers they're using via RA.

Etc. This is just a few examples.

And that's why your argument is frankly morally flawed. Agreements and contracts are something everyone relies on whether they realize it or not. The fix for terms you don't like is not to demand leaky enforcement but, as always, market competition. If you hate Oracle's "no publishing of benchmarks" clauses - which are by the way unenforceable with RA anyway so that's a bad example to pick - then use postgresql instead.

At any rate, it's all rather irrelevant. You have much more bargaining power than you think and that is in fact why RA has never taken off. Outside of very controlled environments like games consoles or cloud VMs it's considered too complicated and excludes too many users who don't have the right hardware etc. Consumers effectively "negotiate" by picking services compatible with what they've got.

you have no idea what you're talking about.

set up the environment properly, so that the attestation can happen, and it will happen.

Are you trying to dispel concerns about remote attestation with "set up the environment properly"? Because that is exactly what has people worried - what is "proper" and who decides it.

Or have I misunderstood you, and you're claiming that Pluton either does not support remote attestation, or allows the owner to fake it? Because as far as I can tell from available information on Pluton and TPM modules, both those claims are false.

Has Microsoft changed its TPM requirements for Windows 11?

If not, then there is no override.

set the registry key and the requirement goes away.

the tpm doesn't hide things from you, it hides things from attackers. if you don't want your disk encrypted, then decrypt it. you can do that. it's very easy.

And now you're straight up lying. The whole purpose of the TPM is to hide cryptographic keys from anyone with physical access, including the owner. There is no override an owner can perform (via, say, a password set on first boot, that may be used to dump all the keys from the TPM into RAM).

It may not by itself force you to encrypt your disk, but it does enable all the abuses listed in https://en.wikipedia.org/wiki/Trusted_Computing#Criticism, that I won't waste summarizing, as it's becoming increasingly obvious you're arguing in bad faith.

The requirement for TPM 2.0 goes away, but Windows 11 still requires TPM 1.2 even with said registry key.

In short, there is no opting out barring an alternative OS.

then turn on the firmware TPM...

> Has Microsoft changed its TPM requirements for Windows 11? > If not, then there is no override.

So going back to my question: The answer is a resounding no.

It’s hard to believe you’re participating in good faith at this point.

Reply on news.ycombinator.com