Data partitioning: an approach to preserving data privacy in computation offload in pervasive computing systems

Offloading computations to remote servers from small mobile devices such as smartphones is a popular technique used in pervasive computing. It addresses the computing and power constraints of small mobile devices. However, a key problem with this technique is a potential loss of data privacy. When a computation is offloaded, user data also needs to be shipped to the possibly untrusted remote nodes. This paper introduces the concept of data partitioning to address this potential loss of data privacy in computation offload to remote nodes. The data partitioning approach allows a user to identify the sensitive parts of her data, which is then prevented from being shipped to untrusted remote servers. The overall execution consists of identifying sensitive parts of user data, shipping code and non-private data for remote execution and getting the results back, and then combining the results from local and remote executions on the mobile device. Data partitioning can be used for a variety of personal digital files that the users create and modify using applications. These include videos, images, audios, and perhaps even textual documents and spreadsheets. It allows mobile users to enjoy a better computing experience by not only further improving the performance and saving power, but also preserving data privacy. The paper demonstrates the applicability of the data partitioning approach via prototypes of three different applications developed for Android devices.