Privacy-Cost Management in Smart Meters with Mutual Information-Based Reinforcement Learning

Abstract

Smart meters (SMs) share fine-grained electricity consumption of households with utility providers almost in realtime. This can violate the users' privacy since sensitive information is leaked through the SMs data. In this study, a novel privacyaware method which exploits the availability of a rechargeable battery (RB) is proposed. It is based on a Markov decision process (MDP) formulation in which the reward received by the agent is designed to control the trade-off between privacy and electricity cost. To obtain a robust and general privacy measure, we adopt the mutual information (MI) between the user's demand load and the masked load seen by the grid. Unlike previous studies, we model the whole temporal correlation in the data to estimate the MI in its general form. The training of the agent is done using a model-free deep reinforcement learning algorithm known as the deep double Q-learning (DDQL) method. In order to estimate the MI-based privacy signal, a neural network termed the Hnetwork is included in the scheme. The performance of the DDQL-MI algorithm is assessed empirically using actual SMs data and compared with simpler privacy measures. The results show significant improvements over the state-of-the-art privacyaware SMs methods.

Real-Time Privacy-Preserving Data Release for Smart Meters

“…Many of the most recent studies of this family, incorporated information theoretic measures such as Mutual Information (MI) or Directed Information (DI) to model the amount of information leaked about the sensitive attributes and used Machine Learning (ML) algorithms for their implementation. On the other hand, the DLS approaches use physical resources such as rechargeable batteries, electric vehicles, and even renewable energy resources, to shape the users' power consumption to mask the sensitive patterns [14]- [25]. Recently, Reinforcement Learning (RL) and Deep Reinforcement Learning (DRL) methods have been used to tackle this problem, showing good performance against strong ML based attackers.…”

On the Impact of Side Information on Smart Meter Privacy-Preserving Methods

“…In recent years, several studies on privacy-preserving approaches for SMs data sharing were conducted, which can be classified into two main families. On the one hand, the methods in the first family [5]- [16] use physical resources such as rechargeable batteries, electric vehicles, heating, ventilation, and air conditioning units, etc., to shape the consumed power so that the SMs measurements reveal minimum information about the user's actual power consumption pattern. The methods in the second family [17]- [24] manipulate the SMs data, to be reported to the utility provider, by distorting it in order to prevent the inference of sensitive information by potential attackers, while preserving the usefulness of the data.…”