I know of CryFS and encFS which both failed an Audit.

I know LUKS and veracrypt which work but are both tedious and not useful for my use case.

I want to encrypt folders on Linux and then sync them with a cloud or just store them locally.

There is goCryptFS which is in Go, i.e. memorysafe which is good. It also failed the audit when it comes to cloud sync.

The issue: if an attacker has access to the encrypted files over time, if you always upload changes, they can crack you.

This would not be relevant for local file stores, but for synced ones it is.

There I used Cryptomator, which has many downsides though

  • the app is GUI only, the CLI variant is last released 2021, which I dont consider maintained when it comes to security? Or is it?
  • the GUI App is in electron, no Wayland support
  • the app devs dont care about flatpak rules and store secrets outside the app container, meaning the app is unconfined by default, you need to manually add an override and it is still pretty insecure in the case that all untrusted apps are flatpaks

KDE KVaults only supports insecure algorithms, goCryptFS would be the most secure for local only stuff, but that was abandoned as a Fedora package and is only available from COPR, currently. (I should learn RPM packaging once again).

Cryptomator seems to be the only one 1. Suitable for cloud sync (not encrypting everything always again) 2. That passed an audit.

The CLI may be fine if it just interacts with the GUI app? I wonder how that would work with the Flatpak.

Do you know any alternatives?

    • pantherina@feddit.orgOPDeutsch
      3·
      21 days ago

      This would only work on single files or archives, but yes probably a good idea. I think cryptomator slices up files into equal pieces and then encrypts both, protecting against analysis of file types, but being sync-friendly.

      • bad_news@lemmy.billiam.net
        1·
        21 days ago

        Probably not an ideal solution, but you could script CLI bcrypt or openssl-ing the files tar-ed and then just rsync or whatever to a location. Organizing by original file hash values would provide unique identifiers that don’t surface the file type without the observer knowing the OG filename and your salt