tang VS mortar

Have a look at clevis and tang. These allow you do have one server - which could be your remote you want to pull from - to be source of the LUKS decryption on the system using tang.

I found that running tang[1] at home and needing to decrypt that box (can be a Pi or whatever) requiring a complex passphrase is very sufficient. You can even just unplug it at night if it makes you sleep better.

https://github.com/latchset/tang

There are 3 ways to unlock a volume in a headless environment: - use a keyfile, located on an already available volume - use your device's TPM and utilize systemd-cryptenroll - use Clevis/Tang to unlock volumes remotely

This is why on headless servers you use tang (ideally, multiple tang servers)

Sharing my notes on running a Tang server on OPNSense, in case it is useful for somebody else.

We use clevis against multiple tang servers to provide Network Bound Disk Encryption (NBDE). It's possible to also use TPM2 but it's easier to use multiple tang servers (requiring more than one server to decrypt) in the datacenter.

For automating unlocking of encrypted drives, look into tang . Here is a red hat guide on setting it up. You will want to be running this on another device on your network, i run it on my router with openwrt since its a local device thats on 24/7. Basically it will unlock your disks as long as your server is on your network, so if your machine or drives are stolen or removed from your network they will just be encrypted as usual. Obviously use a strong encryption password.

There are other ways to bind data, e.g. "network binding" with Tang server.

I wonder if https://github.com/latchset/clevis and https://github.com/latchset/tang (complementary projects) will help here.

Yes you can, using either Mandos or Clevis and Tang. https://www.recompile.se/mandos https://github.com/latchset/clevis https://github.com/latchset/tang. Basically on boot the server gets the key from another(s) servers. You could use a hidden raspberry pi for example.

Bruteforce of such random password is just not plausible and talks about KDF "weakness" is just a distraction. I think most likely it was evil maid attack.

Here are projects which try to mitigate some of evil maid attack risks:

https://github.com/noahbliss/mortar

https://safeboot.dev/

Secure boot and TPM support (à la Mortar: https://github.com/noahbliss/mortar)

There are tools that look to be able to automate it: https://github.com/noahbliss/mortar/blob/master/docs/proxmox-install.md

I just went through the process of setting up new ubuntu VM's using full root disk LUKS encryption and auto-unlock via Proxmox's vTPM2.0 and UEFI ( via this extremely helpful resource https://github.com/noahbliss/mortar )

I have used this project with Debian+proxmox and it's been working great. https://github.com/noahbliss/mortar but I did read the arch wiki a bit which helped my understanding.

There's a project called "mortar" (as in, gluing all these bricks together) that was attempting to simplify this. Though it's lost steam, reading through it's simple bash scripts was a great place to start for me. This guide for Fedora also helped a lot.

There have been a number of attempts to solve this problem, but the most complete appear to be Mortar (a project I head) and safeboot.dev

I highly recommend taking a look at either of these projects if you want be able to improve both your convenience through auto unlocking, and security through broadened scope of audit.

https://github.com/noahbliss/mortar

https://safeboot.dev

Check out mortar. It uses secure boot and TPM along with LUKS. The creator is super helpful and available on the telegram.

A more advanced approach would be something like mortar to chain-load signed stuff.

There seems to be a workable solution out there for 2.0: https://github.com/noahbliss/mortar/blob/master/docs/proxmox-install.md