Owner-Centric Protection of Unstructured Data on Smartphones

摘要: Modern smartphone apps tend to contain and use vast amounts of data that can be broadly classified as structured unstructured. Structured data, such an user's geolocation, has predefined semantics retrieved by well-defined platform APIs. Unstructured on the other hand, relies context reflect its meaning value, is typically provided user directly into app's interface. Recent research shown third-party are leaking highly-sensitive unstructured including banking credentials. Unfortunately, none current solutions focus protection data. In this paper, we propose owner-centric solution protect smartphones. Our approach allows owners specify security policies when providing their untrusted apps. It tracks flow information enforce owner's at strategic exit points. Based approach, design implement a system, called DataChest . We develop several mechanisms reduce burden keep interruption minimum, while same time preventing malicious from tricking user. evaluate our system against set real-world series synthetic attacks show it successfully prevent leakage incurring reasonable performance overhead.