An Information-Theoretic Approach to Time-Series Data Privacy

Access control is central to interfacing with personal data, however most systems today are too coarse and disconnected from the privacy context of the data. Granular access control rarely goes beyond limiting sample rates or enforcing time limits. In this paper, we present a system for tuning a data consumer's access to personal data based on real-time privacy metrics. We first explore the potential definitions of privacy in this context with a focus on information theoretic metrics for defining privacy in sensitive time series data. We then implement and evaluate our system for embedding risk thresholds into bearer token-based access control systems to attenuate access to data of different granularities based on these metrics. Our results show that our system provides privacy gains wihout a significant utility cost, and can run efficiently and scale well on cheap hardware with high-frequncy sensor data.