4 ms·
I don’t agree; Google shouldn’t have made it hard to do this—they should have made it impossible. (And making it merely hard just adds inconvenience while faili
by derefr 8y ago
I don’t agree; Google shouldn’t have made it hard to do this—they should have made it impossible. (And making it merely hard just adds inconvenience while failing at their design goals.)
The point of the Chromebook’s security approach (with the developer mode beep + notification, etc.) is to have a “Trusted Computing Base”, in the original, non-dystopian meaning of the phrase: you’re supposed to be able to tell, as a regular user, just by looking at/booting up a Chromebook, that it’s running Google-signed firmware/software.
A Trusted Computing Base is precisely for the case where you do trust some particular third-party to be your IT admin (because, for example, they’re your school or employer, or—more likely—a subcontractor of one of those that MDM is being delegated to.) In such cases, you want to know that your device is in exactly the state that that party wants it to be in, and that no other party has been able to modify it.
Google aren’t trying to make it impossible for you to use your Chromebook for whatever you want; that’s just a side-effect. Their goal is to make it impossible for someone who intercepts your Chromebook in shipment to install a rootkit on it, in a way where the device still appears to be (and act like) a Chromebook, with no sign that there’s anything wrong.
Yes, Google is the IT admin of the device. Someone’s gotta be; in a Trusted Computing Base scenario, you remove the possibility of handing the thing off to a repair shop (because what is a repair shop—from the TCB perspective—but another attacker?)
Essentially, a Chromebook has the same goal as a high-security lock: nobody should be able to pick it. Not even a locksmith. Not even the locksmith you hired because you lost your keys to the lock you own. Just like with encryption, you can’t leave a backdoor that only works for people with good intentions. You have to leave no backdoor at all—only a “front door.” In the case of a high-security lock, the “front door” is ordering new keys from the manufacturer, supplying your proof of purchase of the lock. In the case of a Chromebook, the “front door” is the signed update system, which allows one party (in this case the manufacturer, but in TCB Windows/macOS PCs this is your enterprise MDM admin) to manage everything. And also spy, if they want. But the two things are one and the same, really. The manufacturer of your high-security lock could break into your building, too. That’s why you pick a lock manufacturer you trust.
Sadly, as the author proves, it’s not impossible to install a rootkit (in this case, a Linux distribution, but it could just as well be a keylogger-injected version of ChromeOS) onto a Chromebook, in a way that’d be invisible to the end-user (unless the end-user X-rays their devices upon receipt and uses ML to compare them to a standard blueprint, ala the brouhaha around spy chips surface-mounted onto the bus of Supermicro servers.)
- deogeo 8y agoUser control is not opposed to security - all your concerns are addressed by making the devices tamper-evident. Making them tamper-proof only benefits Google. As for > you’re supposed to be able to tell, as a regular user, just by looking at/booting up a Chromebook, that it’s running Google-signed firmware/software. Google did do that. But they went a step further - making it extremely easy to accidentally wipe your device and reset it to factory-defaults: https://www.chromium.org/chromium-os/chromiumos-design-docs/developer-mode#_Toc248215938_3369763251394033_7636734712868929 https://www.chromium.org/chromium-os/chromiumos-design-docs/...
- derefr 8y agoYes, and that was the solution to an incentive design problem. If you make it easy to persistently use a rooted device, then regular users would use rooted devices for regular tasks. (Because there are advantages of doing so, like side-loading native apps.) If regular users use rooted devices, some people’s only experience of a device would be of a rooted device. If some people’s only experience of a device is of a rooted device, then they’ll have no idea what an actually tampered device looks like, because every device they’ve seen is “tampered.” This is exactly the state of Android, as it happens. Tons of people use Android in its developer mode; and so people don’t take a security mindset toward using a rooted phone. The solution to this is to use technical mechanisms to achieve social effects (a.k.a. “incentive design” — the type of thing that tax credits are.) By making it irritating and risky to use developer mode in a persistent way, people won’t do it for anything other than tasks that truly require using developer mode for what it’s for (that’s “using the device as a debug build target”, by the way, not “writing code on the device.”) Thus, nobody uses Chromebooks in developer mode. Thus, the average person’s experience of seeing a Chromebook in use is, 99.9% of the time, of a non-rooted Chromebook. Thus, the boot beep is surprising, and will scare the average user off of using the device. Adding a switch in the OS that disables this, would be akin to having a switch in your OS (on a device MDM-managed by someone else, including the certificate store) that disables web browsers from doing any certificate checking. You could flip it; but so could viruses, and so could the NSA half-way through the laptop’s delivery process. Why would you want that switch to exist? —— What people with complaints like yours really want, I think, is not to disable the TCB. You want to be able to be in control of the TCB. You want something that’s like a Chromebook, but where the MDM controller is an enterprise (e.g. you individually, if you’re clever enough) rather than Google. Well, I mean, first: that’s what a regular PC with Secure Boot is. But if the particulars of Chromebooks appeal to you (and I suppose they could appeal to somebody): the cheap hardware, the keyboard layout, the sandboxing, etc. Then what you might want is a variant Chromebook. Such a device would have to look different than a regular Chromebook, somehow, to show regular users that it’s not for them. Sort of like how developer units of game console hardware are clearly not the same as production units. But if that were done, I think it’d be okay for those units to have a user-controllable TCB.