Enhancing user privacy by preprocessing distributed smart meter data

Abstract

The increasing presence of renewable sources requires power grid operators to continuously monitor electricity generation and demand in order to maintain the grid's stability. To this end, smart meters have been deployed to collect real-time information about the current grid load and forward it to the utility in a timely manner. High resolution smart meter data can however reveal the nature of appliances and their mode of operation with high accuracy, and thus endanger user privacy. In this paper, we investigate the impact on user privacy when the consumption data collected by distributed smart metering devices are preprocessed prior to their usage. We therefore assess the impact on the successful classification of appliances when sensor readings are (1) quantized, (2) down-sampled at a lower sampling rate, and (3) averaged by means of an FIR filter. Our evaluation shows that a combination of these preprocessing steps can provide a balanced trade-off that is in the interests of both users (privacy protection) and utilities (near real-time information).

Privacy protection in smart meters using homomorphic encryption: An overview

“…In contrast to CEP, approaches under the change meter data category aim at changing the SM readings before they are used by third-party entities, for example, the utility. There are three main approaches under the CMD category: (1) noise addition (NA), for example, (Barbosa et al, 2014;Fioretto et al, 2019;He et al, 2013); (2) data-downsampling (DS), for example, (Cardenas et al, 2012;Reinhardt et al, 2013;Eibl & Engel, 2015); and (3) data aggregation (DA) (Ilic et al, 2013).…”

“…Approaches via data down-sampling aim at protecting privacy by reducing the temporal frequency at which measurements are available to third-party entities. Typical approaches consist of decimation (consider only every M th sample), filtering (e.g., average and median), and quantization (map the original measurement to a smaller set of discrete finite values; Reinhardt et al, 2013). Despite its simplicity, there is also an evident trade-off between privacy and utility needs when using this approach.…”

Financial impacts of smart meter security and privacy breach

“…One way to do this is by using Nonintrusive Load Monitoring (NILM) algorithms, which are able to distinguish power fluctuations caused by different devices [23], [24]. The power usage information from smart meter can also reveal the private activities of the occupants within the premise such as whether they are awake or asleep, cooking, having meals, taking shower or watching TV [25] - [29]. Inferring these activities is made possible by comparing the energy consumption with the energy profile of household appliances [30].…”

Tensor recovery from noisy and multi-level quantized measurements

“…Due to sensor issues or communication restrictions, images and videos in some applications may have very low resolution [2]. Quantization is applied to enhance the data privacy in power systems and sensor networks [3][4][5]. It is important to develop computationally efficient and reliable methods to recover the actual data from low-resolution measurements.…”