Framework integrating lossy compression and perturbation for the case of smart meter privacy

Abstract

The encoding of high-resolution energy profile datasets from end-users generated by smart electricity meters while maintaining the fidelity of relevant information seems to be one of the backbones of smart electrical markets. In the end-user sphere of smart grids, specific load curves of households can easily be utilized to aggregate detailed information about customer’s daily activities, which would be attractive for cyber attacks. Based on a dataset measured by a smart meter installed in a German household, this paper integrates two complementary approaches to encrypt load profile datasets. On the one hand, the paper explains an integration of a lossy compression and classification technique, which is usable for individual energy consumption profiles of households. On the other hand, a perturbation approach with the Gaussian distribution is used to enhance the safety of a large amount of privacy profiles. By this complete workflow, involving the compression and perturbation, the developed framework sufficiently cut off the chance of de-noising attacks on private data and implement an additional, easy-to-handle layer of data security.