We investigate the extent to which statistical predictive models leak information about their training data. More specifically, based on the use case of household (electrical) energy consumption, we evaluate whether white-box access to auto-regressive (AR) models trained on such data together with background information, such as household energy data aggregates (e.g., monthly billing information) and publicly-available weather data, can lead to inferring fine-grained energy data of any particular household. We construct two adversarial models aiming to infer fine-grained energy consumption patterns. Both threat models use monthly billing information of target households. The second adversary has access to the AR model for a cluster of households containing the target household. Using two real-world energy datasets, we demonstrate that this adversary can apply maximuma posterioriestimation to reconstruct daily consumption of target households with significantly lower error than the first adversary, which serves as a baseline. Such fine-grained data can essentially expose private information, such as occupancy levels. Finally, we use differential privacy (DP) to alleviate the privacy concerns of the adversary in dis-aggregating energy data. Our evaluations show that differentially private model parameters offer strong privacy protection against the adversary with moderate utility, captured in terms of model fitness to the cluster.

Smart meter data is collected and shared with different stakeholders involved in a smart grid ecosystem. The fine-grained energy data is extremely useful for grid operations and maintenance, monitoring and for market segmentation purposes. However, sharing and releasing fine-grained energy data induces explicit violations of private information of consumers (Molina-Markham et al., 2010). Service providers do then share and release aggregated statistics to preserve the privacy of consumers with data aggregation aiming at reducing the risks of individual consumption traces being revealed. In this paper, we show that an adversary can reconstruct individual traces of energy data by exploiting consistency (similar consumption patterns over time) and distinctiveness (one household’s energy consumption pattern is significantly different from that of others) properties of individual consumption load patterns. We propose an unsupervised attack framework to recover hourly energy consumption ti me-series of individual users without any prior knowledge. We pose the problem of assigning aggregated energy consumption meter readings to individuals as an assignment problem and solve it by the Hungarian algorithm (Xu et al., 2017; Kuhn, 1955). Using two real-world datasets, our empirical evaluations show that an adversary is capable of recovering over 70% of households’ energy consumption patterns with over 90% accuracy.