Guarding Sensitive Sensor Data against Malicious Mobile Applications

With increasing usage of sensor data for medical purposes, the ability to secure sensitive features in mobile sensor data from adversarial applications is a continuous challenge. This paper introduces a random anonymization algorithm, SparCTym, as a method for anonymizing sensitive features in walking accelerometer data while maintaining the utility of the data. SparCTym was implemented in the Android framework of a Nexus S phone and tested with activity recognition applications.