BioMyth (He/Him)

Like others are saying, a simple fix to this is to setup the homeassistant machine for https & a self signed cert. Then on the Caddy machine you can configure the https to not verify the origin. That would make the communications more robust, but I think it is still vulnerable to MITM attacks.

While this is a risk, it is only a real risk if the system is already exploited for regular user access. Or if there is an untrustworthy user of the system. So for most, it is not a major concern.

I have determined that foot is best for me personally, like alacritty and a couple others, it is very barebones. No tabs or anything like that without tmux. But it doesn’t rely on GPU acceleration and is just as fast (or faster) than my experience using GPU accelerated terminals. Easy to configure and since it doesn’t have the GPU requirements it works on old hardware like a dream. Only possible issue is that it is wayland only but since that is all I like to use it is perfect.

I find a lot like ghostty and wezterm try to include too many features. All I need a terminal emulator to be is a terminal emulator. But then a lot of these then add tabs, build in multiplexers & more and it is more bloated than I like a simple utility to be. Additionally, I don’t need native tabs as a lot I do in the terminal uses SSH so it is easier just to use tmux/zilji and not have to manage it as much.

It is the largest reason. Storing the password is one thing but to make the device reasonable to use I would likely store the key’s in TPM with a backup key. I don’t think she would be technical enough to use the backup keys were something additional to happen.

BioMyth

I understand that giving the keys can partially solve the access problem. But she would still possibly be unable to use the device. Additionally, I don’t know that she would be capable of using the keys without additional assistance and we don’t have other techies in our community who could step up in that capacity.

I don’t for a pretty simple reason. I have a wife, if something ever happened to me then she could end up a creek without a paddle. So by not having it encrypted then, anyone kinda technical can just pull data off the drive.

I’m on the bandwagon of not hosting it myself. It really breaks down to a level of commitment & surface area issue for me.

Commitment: I know my server OS isn’t setup as well as it could be for mission critical software/uptime. I’m a hobbiest with limited time to spend on this hobby and I can’t spend 100hrs getting it all right.

Surface Area: I host a bunch of non mission critical services on one server and if I was hosting a password manager it would also be on that server. So I have a very large attack surface area and a weakness in one of those could result in all my passwords & more stored in the manager being exposed.

So I don’t trust my own OS to be fully secure and I don’t trust the other services and my configurations of them to be secure either. Given that any compromise of my password manager would be devastating. I let someone else host it.

I’ve seen that in the occassional cases when password managers have been compromised, the attacker only ends up with non encrypted user data & encrypted passwords. The encrypted passwords are practically unbreakable. The services also hire professionals who host and work in hosting for a living. And usually have better data siloing than I can afford.

All that to say I use bitwarden. It is an open source system which has plenty of security built into the model so even if compromised I don’t think my passwords are at risk. And I believe they are more well equipped to ensure that data is being managed well.

OpenSUSE tumbleweed is a good compromise IMO. it is also a rolling release distro with built in snapshotting. So if anything does go wrong it takes ~5 mins to roll back to the last good snapshot. You can set the same thing up on arch but it isn’t ootb and YAST is a great management tool as well.