
Encryption specs are are here: https://spideroak.com/engineering_matters#encryption_specifications
We like to say that trust isn't necessary because we're incapable of betraying our users. It's makes good business sense too. We don't want to spend our time answering subpoenas.
To add your own layer of encryption, you can archive container files or whatever you like. No limits. If you a sector based encrypted disk image, SpiderOak will be able to efficiently snapshot it between versions, giving you history and only saving the changes between revisions.
If you want a layer of additional local control, there's a "Keep your own copy" option where SpiderOak will put a copy of every encrypted data block on your own server, so you can manually inspect them if you wish (and have offline/local access for very fast restores.)
There's no reason online can't be secure. Online means it's automatically offsite and that a 3rd party has the time and incentive to be sure it's actually working.
2 years ago I founded https://spideroak.com/ for this exact situation -- wanting a zero-knowledge approach to encryption. We explicitly don't know anything about your data. We just see boring sequentially numbered data blocks on the server. Instead of a EULA, we have a "remember your password" agreement.
You can combine data from unlimited devices and it de-duplicates, and can automatically sync folders for you. Storage is perpetual (unless you explicitly remove things.) FWIW, it's written in Python and we have always supported Linux.
"You need tender loving care once a week - so that I can slap you into shape." - Ellyn Mustard