In smart-metered systems, fine-grained power demand data (load profile) is communicated from a user to the utility provider. The correlation of the load profile with a user's private activities leaves open the possibility of inference attacks. Using a rechargeable battery, the user can partially obscure its load profile and provide some protection to the private information using various battery charging policies. Using the mutual information as the privacy metric we study a class of optimal battery charging policies.
When the power demand is a first-order Markov process, we propose a series of reductions on the optimization problem and ultimately recast it as a Markov decision process. In the special case of i.i.d. demand, we explicitly characterize the optimal policy and show that the associated privacy-leakage can be expressed as a single-letter mutual information expression.
Published December 2015 , 23 pages