Platform-level protection for interacting mobile apps

Xu, Yuanzhong, Ph.D.
Journal Title
Journal ISSN
Volume Title

In a modern mobile platform, apps are mutually distrustful, but they share the same device and frequently interact with each other. This dissertation shows how existing platforms, like Android and iOS, often fail to support important data protection scenarios, and describes two systems to improve platform-level security.

First, many data leaks in existing platforms are due to the lack of information flow control for inter-app data exchanges. For example, a document viewer that opens an attachment from an email client often further discloses the attachment to other apps or to the network. To prevent such leaks, we need strict information flow confinement, but a challenge to enforce such confinement in existing platforms is the potential disruptions to confined apps. We present Maxoid, a system that uses context-aware custom views of apps' storage state to make information flow enforcement backward compatible.

Second, apps' abstraction of data has diverged from platforms' abstraction of data. Modern mobile apps heavily rely on structured data, and relational databases have become the hub for apps' internal data management. However, in existing platforms, protection mechanisms are coarse-grained and have no visibility to the structures of apps' data. In these platforms, access control is a mixture of coarse-grained mechanisms and many ad hoc user-level checks, making data protection unprincipled and error-prone. We present Earp, a new mobile platform that combines simple object-level permissions and capability relationships among objects to naturally protect structured data for mobile apps. It achieves a uniform abstraction for storing, sharing and efficiently protecting structured data, for both storage and inter-app services.