Paper 2012/183

Differentially Private Smart Metering with Battery Recharging

Michael Backes and Sebastian Meiser

Abstract

The energy industry has recently begun using smart meters to take fine-grained readings of energy usage. These smart meters enable flexible time-of-use billing, forecasting, and demand response, but they also raise serious user privacy concerns. We propose a novel technique for provably hiding sensitive power consumption information in the overall power consumption stream. Our technique relies on a rechargeable battery that is connected to the household's power supply. This battery is used to modify the household's power consumption by adding or subtracting noise (i.e., increasing or decreasing power consumption), in order to establish strong privacy guarantees in the sense of differential privacy. To achieve these privacy guarantees in realistic settings, we first investigate the influence of, and the interplay between, capacity and throughput bounds that batteries face in reality. We then propose an integrated method based on noise cascading that allows for recharging the battery on-the-fly so that differential privacy is retained, while adhering to capacity and throughput constraints, and while keeping the additional consumption of energy induced by our technique to a minimum.