In recent years wireless technology has allowed the power of lightweight (thermal, light, motion, etc.) sensors to be explored. This data offers important benefits to society. For example, thermal sensor data now plays an important role in controlling HVAC systems and minimising energy consumption in smart buildings. Simultaneously, we have begun to understand the extent to which our privacy is compromised by allowing this increased level of data collection. In particular, allowing sensors into the home has resulted in considerable privacy concerns. The field of privacy-preserving data analytics has developed to help alleviate these privacy concerns. A particular notion of privacy, which will be our focus, called differential privacy has emerged as a gold standard for privacy.

In this talk we will define ``local differential privacy" and discuss the utility of locally differentially private thermal sensor data. In particular, we’ll discuss results that indicate that we can produce a version of the sensor data that keeps the exact locations of a heat source private, while allowing a data analyst to determine the general geographic vicinity of the heat source. If we have time, we’ll also touch on general linear inverse problems. We'll briefly discuss the connection between the how ``easy” it is to preserve privacy and the condition number of a matrix. Speaker(s): Audra McMillan (University of Michigan)