Proper sandboxing and effective security permissions. Third party code is inherently still a black box, especially when the bulk of the would be attacker's code

Proper sandboxing and effective security permissions. Third party code is inherently still a black box, especially when the bulk of the would be attacker's code is running on a server and never goes through Apple's review process. There is no way to tell the difference between an app that uploads all of a user's contacts to do a one time search, and one that backhauls and stores all those contacts for later nefarious purposes. (I believe) Apple came up with a rule to directly prohibit mass uploads to address this specific thing, but the cat and mouse dynamic remains. Ultimately users need to be able to deny apps permissions, based on what an app needs to do and how much a user wants to trust a company (by default, very little). Permissions need to be implemented such that apps are unaware they're being denied so they're unable to coerce users into enabling them. For example, "location access" needs to have the option of returning a virtual location that was set from a map.