Towards provable privacy guarantees using rechargeable energy-storage devices

The global energy transition requires the availability of energy-consumption data with high resolution. Smart meters record such data in real time. This however endangers privacy: Time series of energy-consumption data contain different kinds of private information, such as the employment status of the residents. We address this problem by proposing a consumption-perturbation approach that relies on energy-storage devices (aka. batteries). The energy (dis-)charged to them perturbs the actual data describing the consumption. So-called charging strategies specify the (dis-)charging behavior. A main objective of this article is to come up with privacy guarantees for such strategies. To this end, the strategies we propose rely on a generalization of the Irwin-Hall distribution, which facilitates closed-form analyses. For these strategies, we derive (∈, δ)-differential privacy guarantees. Next, we propose a new measure, which is statistical in nature, to quantify the risk of confusing the assignment of features to the time series they are computed on. We then develop a specific charging strategy that combines the properties required to provide the guarantees proven earlier with trend preservation to shield against filtering approaches. All in all, our strategies increase the failure probability of approaches inferring private information from the data.