文章基本信息

标题：A Fundamental Performance Limit of Cloud-based Control in Terms of Differential Privacy Level ⁎
本地全文：下载
作者：Yu Kawano ; Kenji Kashima ; Ming Cao 等
期刊名称：IFAC PapersOnLine
印刷版ISSN：2405-8963
出版年度：2020
卷号：53
期号：2
页码：11018-11023
DOI：10.1016/j.ifacol.2020.12.220
语种：English
出版社：Elsevier
摘要：AbstractIn this paper, we address a privacy issue raised by cloud based control. In a cloud based control framework, a plant typically has no access to the models of the cloud system and other plants connected via the cloud system. Under restricted information, the plant is required to design its local controller for achieving control objectives. As a control objective, we consider a tracking problem, and for constant reference signals, a class of tracking controllers is identified based on Youla parametrization. More importantly, as local tracking controllers are implemented, there is a possibility that the cloud system or other plants connected via the cloud system may be able to identify private information of the plant by using the collected signal from the plant; for example, the reference signal (say, the target production amount) of the plant can be viewed as a piece of private information. In order to evaluate the privacy level of the reference signal, we employ the concept of differential privacy. For the Laplace mechanism induced by the entire system, we show that the differential privacy level cannot be further improved from a ceiling value for any parameters of the local controller. In other words, there is a performance limit in terms of differential privacy level, which is determined by the plant and cloud system only.
关键词：KeywordsDifferential privacyprivacy limitcloud-based controldiscrete-time linear systems