Information Theoretic Privacy in Electric Grids with Smart Meters

Ashish Khisti
Associate Professor, University of Toronto
Date: Oct. 28th, 2016

Abstract

Smart meters report electricity usage of a user to the utility provider on a real-time basis, which is known to leak sensitive user information. In this talk we will discuss how a rechargeable battery with limited storage capacity at the user’s home can be used to mask this information.

We will use the mutual information between the user load and the grid output as our privacy metric and assume that the rechargeable battery satisfies ideal charge conservation. We show that the problem of designing optimal charging policies is equivalent to designing a communication channel subject to certain state constraints. For the case of i.i.d. inputs we derive an explicit solution and provide an intuitive interpretation based on certain invariance properties of the system. For the case of Markov inputs we cast the problem as a Markov Decision Process (MDP) that could be solved using a dynamic program. We will also discuss a generalization when multiple batteries cascaded in series can be used by the system.

This is a joint work with Simon Li and Aditya Mahajan.

Bio

Ashish Khisti is an Associate Professor and a Canada Research Chair in the Dept. of Electrical and Computer Engineering at the University of Toronto, Canada. He was an assistant professor at the same university between 2009-2015. He received his BASc degree from the Engineering Science Program, also from University of Toronto, and his MASc and PhD degrees in EECS from the Massachusetts Institute of Technology. His present research interests include information theoretic security & privacy, and real-time communication systems. He is presently a consultant for Cisco Systems.