eduzhai > Applied Sciences > Engineering >

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

  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: Smart meters (SMs) can pose privacy threats for consumers, an issue that hasreceived significant attention in recent years. This paper studies the impactof Side Information (SI) on the performance of distortion-based real-timeprivacy-preserving algorithms for SMs. In particular, we consider a deepadversarial learning framework, in which the desired releaser (a recurrentneural network) is trained by fighting against an adversary network untilconvergence. To define the loss functions, two different approaches areconsidered: the Causal Adversarial Learning (CAL) and the Directed Information(DI)-based learning. The main difference between these approaches is in how theprivacy term is measured during the training process. On the one hand, thereleaser in the CAL method, by getting supervision from the actual values ofthe private variables and feedback from the adversary performance, tries tominimize the adversary log-likelihood. On the other hand, the releaser in theDI approach completely relies on the feedback received from the adversary andis optimized to maximize its uncertainty. The performance of these twoalgorithms is evaluated empirically using real-world SMs data, considering anattacker with access to SI (e.g., the day of the week) that tries to infer theoccupancy status from the released SMs data. The results show that, althoughthey perform similarly when the attacker does not exploit the SI, in general,the CAL method is less sensitive to the inclusion of SI. However, in bothcases, privacy levels are significantly affected, particularly when multiplesources of SI are included.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×