So it was/is recommended to use a pin/key and/or recovery key to ensure the security of the data. Unless your only threat model was to protect against common thievery and assume the attack had no technical prowess (and that’s perfectly fine, I do this for my company). Not to mention they were kinda used as a warranty canary for Truecrypt [2]. There were suspicions that nation states may have hardware bypasses worked out.
Later there were implementations of hardware encryption found to be vulnerable. So even now bitlicker does everything in software by default. [3]
So I understand why FOSS devs would rely more on standard practice (shared keys) with LUKS and not embrace hardware enclave options like TPM. They haven’t been the most reliable over the long term and are harder to patch/fix.
[1] https://pulsesecurity.co.nz/articles/TPM-sniffing
[2] https://threatpost.com/of-truecrypt-and-warrant-canaries/106...
[3] https://www.technadu.com/bitlocker-to-use-software-encryptio...
Bitlocker is still vulnerable to key exfiltration attacks because it's not using any encrypted communication protocols that exist in the TPM standard, but that can be (and should be!) fixed.
In the end, I use encryption to make sure nobody can just plug in a flash drive and copy all my personal files and passwords off my laptop. If they have the time and tools to exfiltrate the security key through the SPI bus, they probably have the means to install a hardware key logger in my keyboard as well. The attacks against TPMs are out of scope for my threat model and honestly they probably should be for anyone but businesses carrying secrets as well.
If the United States or China wanted my passwords that badly, they'd probably just drug me or hit me until I hand them over. Defending against such adversaries requires more than just encryption, you'd need to use something like Qubes and alter your entire lifestyle to be secure.
What I want is to have a Linux system where I can turn it on without a password and have a good reason to believe that my files weren't compromises by the maid and that the OS didn't get keylogged. That requires several parts working together.
Getting secure boot to work is easy enough these days, but once you get through secure boot you're in for a challenge. I don't know of any stable bootloaders that don't allow you to edit the init binary to /bin/bash to give you a root shell from the menu, which is a requirement for the ease of use Windows provides. I also don't know if it's even possible to get a chain of trust from initramfs back to the hardware like Windows allows for. The *BSDs seem to be doing some kind of checksumming, but I don't know how far Linux is along with this.
In my ideal world, you get prompted on how to encrypt your Linux system upon install. "Disabled", "Automatic" or "Secure", with a note that "secure" is probably what you want if you can't pick but you have to provide a password at boot. I'd also like for popular distros to switch to full disk encryption because the unencrypted boot partition defeats half the point without secure boot and custom keys (which nobody actually uses).
Microsoft proves that this can be done, although their default allows for booting without a password a bit too easy. If you buy Windows with a Pro key, they'll encrypt your system in place with the click of a button. Everyone can set it up, and in many cases it's even the default. This is a basic usability security feature that Linux just can't compete with, and in my opinion that's a shame.
> modern chips are a lot more reselient against crashes
sorry for a basic question, but im not sure how a chip itself is more resistant to a crash (in os? user space?)...I think it has something to do with the modern instruction sets being kinder to the kernel and the fact that on computers with recent processors certain processor features are enabled in the UEFI config by default more often, but I couldn't tell you which features that would be. My hunch is that I has to to with stuff like virtualisation based security and the like?
If a program crashes on an older CPU, it damn sure will crash on a modern CPU just as well.
Not necessarily; if the crash is caused by an instruction that's absent on older CPUs (for instance, trying to use an AVX2 instruction when the most the CPU has is SSE2), it will work on a modern CPU but crash on an older CPU.
If foo can be launched on a platform that it’s not meant for, it’s an issue with foo, not the platform.
Why would you present Microsoft PR as fact?
Sort of a side point, but this got me wondering...Is there something inherently less stable about these older chips, or maybe is their stability somehow a function of their lifetime that would really matter here? My own anecdata (which is from a far smaller dataset than what I imagine Microsoft would have access to) would suggest that this isn't really the case, at least for anything otherwise capable of running something like Windows 10 or Windows 11, but I'd be interested in reading more about it.
Is it that old systems tend to not be physically maintained as well thus resulting in cooling issues and more overheating?
Is it that these "crashes" are application crashes due to the attempted execution of instructions in (newer) x86 extensions not implemented by these (older) chips?
Whether it's because only newer chips without wear and tear come out on top in these statistics or because there's something in the hardware itself, the perception that Windows 11 is more stable is something Microsoft can market. Dropping chips that lack certain instructions also make their support and testing workload lighter. In the end, the quality and range of support Microsoft provide for their operating system depends on how much money they can make off their sales. If their losses from the move are lower than the cost of supporting older hardware, it's a decent business decision to do the unpopular thing and drop support. It's a private company, after all, focused on making their shareholders money.
There's another part to the exclusion of old hardware, which is that modern chips are a lot more reselient against crashes according to the telemetry Microsoft collects. The same is true for secure boot and other security lockdowns every Linux user disables. You could make the argument that this means that Microsoft is failing to provide stability for this older hardware, but it doesn't necessarily mean that it makes business sense for MS to put money and resources towards resolving the issue. Not making Windows 11 available on old chips doesn't hurt sales, helps them boast with great stability and security statistics and barely makes a dent in their reputation. Most people with a negative opinion of the company here were hating on Microsoft long before Windows 11 was even announced.
The TPM story makes sense from a Windows Hello standpoint. I don't think there's any doubt that the hardware trust system is more secure than the previous system. However, that trust is completely useless because Microsoft STILL doesn't enable Bitlocker unless you pay extra. It's current_year and Microsoft still hasn't brought data security to the masses. This is an area where proper use of the TPM can be benefit users massively.
Linux is having the exact opposite problem, I want to use my TPM and secure boot to leverage the hardware security built into my devices but it's as if every part of the Linux boot chain has implemented some kind of limitation to make the process difficult. Bitlocker works great, and I want it on Linux too, but nobody writing code for the Linux ecosystem seems to share my preferences here.