The peace of mind in having all your sensitive data under your control is totally worth it.
I used to have some illusions that "if I self host, I am in control", and "if I don't connect my home infra to the internet, I am safe". Later I realized neither is true.
I can't trust all the consumer grade devices in my network, I don't trust a software just because it is open source. And I don't have time to keep up with all the security patches and do security auditing / vulnerability scan routinely...etc.
It is fine to self host hobby stuff for fun, but professionally managing sensitive data is a full time job.
If you want to really get paranoid, pass it all through wireguard or ssh tunnels, but for bitwarden at least it's all client side encrypted anyways, you could probably run it on a very out of date system without issue.
And these are just for the integrity of your _encrypted_ data. There are a lot more to do to fully secure your home infra in general. How do you secure your wireguard client key on the go? Do you monitor access logs? What about Guest Wi-Fi access, vlan separation...
I don't know if worrying about all these considered being paranoid. End of day it's about risk management, and personally, the benefits of selfhosting does not justify the effort I will need to put into maintaining it.
The server does not see the domains you have passwords for. The following data are saved in plaintext:
- A list of "equivalent domains" (this starts out with a default list, but individuals can change this). This totally can be used to deduce which websites you have an account for, but that's not really enough information, as most websites will not have an entry here.
- Some metadata such as your email, master password hint
- Most of the boolean values (mfa enabled, email verified, premium)
- Custom field types (types only, field name, and value are both encrypted)
- Revision date
- Bunch of UUIDs
Here is what a single password entry looks like when retrieved from the /sync endpoint, which happens before decryption: https://pastebin.com/FLr19qiN
> You have to trust the server also if you use the web client because the web client is loaded from the server.
This is true! However, the android app, cli, and other clients do not get loaded from the server, thus, in theory, you can inspect the source of them, possibly compile it yourself, and use that. In those scenarios you do not have to trust the server.
The situation with Bitwarden is a bit different though. Secrets are encrypted on the clients, the server never sees decrypted data.
An attacker can justify allocating a lot of resource to pwn bitwarden.com. If you manage to break into the vaults you're bound to find something juicy, just because of how large the target is.
Or you could decide to pwn me. Figure out where my bitwarden is hosted, what my config looks like, what mistakes I might have made setting it up, then maybe find a way in. Then it's just the start, since the passwords are encrypted on disk, so at best you have access to an encrypted sqlite database. Now you need to find a way to get me to leak my vault key. Maybe I sometimes use the web interface? Maybe not. Who knows.
After all of that you may realize that all of my passwords are either not super important or require some form of 2FA, therefore wasting your time.
But hey, you can log into my hacker news account!
Security through irrelevancy.
If you're going to host services as home such as your password manager, set up a WireGuard VPN, you can use a Pi and it'll be perfectly sufficient, leave only the VPN open on the internet, VPN in from your phone, laptop, whatever for anything you need access to, and you don't need to rely on Nextcloud or Bitwarden having vulnerabilities discovered in them.
I was using Nextcloud previously for password sync because my password manager needs WebDAV, it was too much to maintain so I wrote a small server in Golang using the WebDAV library and it sits behind NGINX which handles the auth. I run Minio (S3 compatible) for syncing our family photos from our phones and Folder Sync app on Android. They both run on a VM and write out to a ZFS pool.
I have a Pi 3B+ running Raspbian mounted read-only as a WireGuard VPN for remote access, and we use the official WireGuard app. VPN is always on because we have fast, symmetric fibre, and we don't need to worry about trusting public networks.
I've had my Raspberry Pis kill dozens of SD cards over the years, so read only can helas for updates, I manually remount read/write when I do maintenance and then remount read/only again when I'm done.
Switched to Bitwarden instead.
Personally I think hosting the server locally doesn't give much benefit because I'm more likely to screw things up than Bitwarden is on that front.