Something that I've found myself thinking a lot about recently is the difficulty of ensuring that the code/app/software a user is running (on their own device) hasn't been replaced with a compromised/backdoored/trojan'd doppelganger that's secretly siphoning off all your data.
It's effectively a DRM challenge, and the sort of thing you can handle using a secret if the platform is secure but it's not clear to me whether solutions like ARM's TrustZone can be leveraged to protect secrets in this way.