Hi! I want to try out fedora workstation in the near future (once 39 is out) and was wondering if systemd-homed is ready for everyday use yet.
I’m a bit paranoid and really need my private data encrypted. However, I don’t think that full disk encryption is practical for my daily use. Therefore I was really looking forward to the encryption possibilities of systemd-homed.
However, after reading up on it, I was a bit discouraged. AFAIK, there’s no option to setup systemd-homed at installation (of fedora). I was an Arch then Manjaro, then Endeavour user for years but don’t have the time/patience anymore to configure major parrts of my system anymore. Also, the documentation doesn’t seem too noob-friendly to me, which also plays into the time/patience argument.
Is it ready? Can anyone seriously recommend it for a lazy ex-Arch user who doesn’t want to break another linux installation?
Thank you in advance. :)
I second this.
Full disk encryption is entirely practical for everyday use. If you don’t already have a dedicated TPM, your motherboard/CPU may provide a software TPM (fTPM?). If so, you don’t even have to interact with the machine during boot. It’s just a bit slower to start up (by a few seconds), which really isn’t a big issue for your average user.
Pardon my ignorance here, but I don’t get it how is the whole thing still safe with unlocking from TPM instead of me providing the password at boot time?
Considering now anyone can just boot the machine into the installed system then bruteforce/exploit something to get login/get read permissions and make a plain copy of the data?
Where, without tpm, as long as I do not type in the encryption password myself I have a pretty high guarantee that the data is safe, especially when I am not at the (powered down) computer.
This is what I don’t understand either. It seems like with tpm it only protects the data from someone taking or copying the hard drive, but the bigger risk seems like what you describe
plus, using an encryption password and then automatically logging in the user prevents needing to enter two passwords while still keeping the data secure as long as the machine is off
You have a competent grasp of the situation
The idea behind it is that the files are stored encrypted at rest, which is really what you want, because once a system is booted, you have to play by the computer’s rules (respect file permissions, policies, etc.).
The TPM provides a secure mechanism to provide a decryption key to the computer during boot, eliminating the need for direct interaction.
Could it be compromised? Probably, but it would take considerably more effort than a man-in-the-middle on your keyboard via a logic analyzer.
This is a common misunderstanding insofar as how encryption works. You can’t flick a bit and TURN your storage unencrypted nor can you plausibly make your computer obey restrictions.
If your storage is encrypted it remains encrypted always including the file you have open right now. Your takes a plausibly length usable string and uses it to compute or retrieve the long binary number actually needed to decrypt your files. This number is stored in memory such that encrypted files can be decrypted when read into memory.
Once that key is loaded in memory anyone with 10 minutes and access to google could trivially unlock your computer in several different ways. It is virtually exactly like having no security whatsoever.
If you don’t actually enter a passphrase to unlock you have no meaningful security against anything but the most casual unmotivated snooping.
Your little sister might not be motivated enough to read your diary but the dipstick that stole your laptop will definitely be spending your money.
I highly doubt it.
If you have any tips for how I can personally bypass my computer’s encryption in 10 minutes without being able to login, I’d love to try my hand at it.
You aren’t actually asking to how to bypass encryption because the key is already in memory. You are asking about the much simpler task of compromising a computer with physical access to same. Depending on configuration this can be as ridiculous as killing the lockscreen process or as hard as physically opening the case chilling the contents of ram enough that data survives transfer to different physical hardware. See also the massive attack surface of the USB stack.
That doesn’t sound trivial at all.
On most systems you can press a hotkey in grub to edit the Linux command line that will be booted and in about 7 keystrokes gain access to any unlocked filesystem. Asking how you can break into a system you physically control is like asking how many ways you could break into a house supposing you had an hour alone with a crowbar the answers are legion. No machine in someone else’s hand which is unlocked can possibly be deemed secure.
Even dumber no installer will create such an insecure configuration because the people that design Linux installers are smarter than you.
I’m not advocating for this right now, but yes that is why when using TPM password, one must insure to enable secure boot, enable bios password, disable boot media, and disable grub editing. That’s the recommended proceedure for this setup.
This is essentially how HEADs works too. Some very smart people have worked on TPM boot and it is even built into systemd. You’re just wrong here.
Reference:
Whether I would fully rely on the systems proper operation against a state sponsored adversary is a different question though.
Actually, thinking more about this…
Can you give an example of this grub cmdline bypass? If what you’re saying is true, this would be a huge issue. I’d switch bootloaders over something like this.
Though after a point rubber hose cryptanalysis will become the more pragmatic option for an attacker.
deleted by creator
How so? The data is still encrypted on the drive after boot, so unless your machine also automatically logs you in, there shouldn’t be anything to worry about.
deleted by creator
Very true, which is why it’s important to run as few services and have a locked down firewall. Maintaining a minimal attack surface is everything.
If the attack involves stealing the machine it has been demonstrated that you can ice the ram and swap it into another machine without data loss, recover the keys, and access the drive.
If we’re talking apt adversaries it’s also possible to just build a ram sniffer
While true, I think most people’s concern is that their laptop is stolen and along with it all the access details for their email, online banking and so on.
If you’re doing things that mean you’re going to be the target of people with the knowledge, time, and technology to freeze the RAM and attempt to recover the data, you’re presumably already well aware of those (and other) dangers anyway.
Those people need to learn how to fend of these attacks somewhere, too.
I guess my point was that if you’re going to use FDE and unlock the disk without a password, don’t bother hardening against attacks that involve stealing the whole machine.
Agreed that this is almost fiction level paranoia.
As others explained: If the FDE key is in RAM, I’m vulnerable. My thread model includes a stolen Laptop with the attackers able to freeze my RAM and reading out the keys.
Thank you for mentioning TPM though. Didn’t know of that before. :)
There are plenty of reasons to not want FDE or not want just FDE alone. Shared computer, your data isn’t safe if you share the FDE password with another user who needs to share the system. He said he’s paranoid, so he is wanting his data encrypted above all. Home directory encryption, especially on top of FDE, while a performance hit, would do well for that. But most importantly, he said FDE isn’t practical for him, end of FDE story.
That’s a very absolutist way to look at a situation. It’s equally likely (in fact, much more likely) that OP is missing a detail or two about FDE, and we won’t know for sure until we discuss it.
The question was specific to systemd-homed. Jumping to why isn’t FDE good enough for you isn’t even logical. Sure they changed direction later, but it is not what was originally asked. In answering the primary question, additional questions from the OP may have arisen, which is fine. What is not fine is to assume incompetence from the start, which is what ya’ll were doing.