Privacy Capsules: Preventing Information Leaks by Mobile Apps.

ABSTRACTPreventing the leakage of user information via untrusted third-party apps is a key challenge in mobile privacy. We propose and evaluate privacy capsules (PCs), a platform execution model for mobile apps that prevents the flow of private information to untrusted parties by design. With PCs, apps execute in two sequential phases. In the unsealed phase, the app has no access to sensitive input but full access to untrusted network resources. In the sealed state, the untrusted app has access to sensitive input, but can no longer communicate with untrusted resources. Privacy capsules are implemented by the mobile platform, are language independent, and require few changes to apps. Using a prototype PC implementation in Android, we show that PCs have low performance and energy overhead, and are suitable for a large class of apps.