Pervasive computing, such as sensors in smartphones, buildings, automobiles and cities, result in increased sharing of sensor data, whether initiated by users or by other authorities such as service providers, government entities, interest groups, and individuals. Embedded in this data is information which others, even using sophisticated data mining algorithms, can fuse to construct a virtual biography of our activities, revealing private behaviors and lifestyle patterns. Researchers in this project are devising computational methods to let users exercise privacy control over their personal sensory data that is shared.
Intellectual Merit: The project is developing a user-configurable cryptographically-secure ?privacy shield? to run on smartphones and act upon sensor information flowing to other users, apps, and services. To make privacy understandable, the user is presented with a higher level abstraction for expressing privacy and sharing in terms of rich inferences and contexts drawn from sensor measurements. The user can designate some inferences and contexts as private. To provide privacy while ensuring the quality of service provided by the recipients of the sensory information, the system also incorporates algorithms which, over time, learn a personalized model of the privacy risk from sharing an inference. The theoretical concepts and the system realization are being validated via user studies in mobile health and personal sensing.
Broader Impacts: By providing better understanding of the behavioral privacy problem and risks inherent in sharing seemingly innocuous data, results from this project will lead to a more educated and informed citizenry, regulators, and policy makers, and provide effective tools for privacy management to those who share sensory information.
