PhD Candidate, University of California, Berkeley
Towards a Theory of Free-Lunch Privacy in Cyber-Physical Systems
One of the hallmarks of Cyber-Physical Systems (CPS), ranging from smart homes, smart transportation systems, smart energy systems, to smart cities, is that data collected from individuals or entities serve an indispensable part of decision making and control underlying the systems’ operation. There has been a long-standing argument as to the tradeoff between privacy and data utility. In this work, we adopt a multiparameteric programming approach to rigorously study conditions under which data utility has to be sacrificed to protect privacy and situations where free-lunch privacy can be achieved, i.e., data can be concealed without hurting the optimality of the decision making underlying the CPS. We formalize the concept of free-lunch privacy, and establish various results on its existence, geometry, as well as efficient computation methods. We propose the free-lunch privacy mechanism, which is a pragmatic mechanism that exploits free-lunch privacy if it exists with the constant guarantee of optimal usage of data. We study the resilience of this mechanism against attacks that attempt to infer the parameter of a user’s data generating process. We demonstrate the efficacy of the mechanism via a case study on occupancy-adaptive smart home temperature control.
Ruoxi Jia received a Bachelor of Science degree in EECS (2013) from Peking University. She is currently pursuing her PhD in the Department of EECS at the University of California, Berkeley. She also works in the Building Technology and Urban Systems Division in the Lawrence Berkeley National Lab. Ruoxi is a winner of The Chiang Fellowship for Graduate Scholars in Manufacturing and Engineering at UC Berkeley. She is also an awardee of the 2108 Alumni Scholarship and the Okamatsu Scholarship. Her current research interests are at the intersection of control, optimization, statistics, cyber-physical system and smart buildings. Ruoxi’s dissertation research focuses on improving efficiency, intelligence, privacy of cyber-physical systems. Ruoxi’s work has been featured in the MIT Technology Review and IEEE Spectrum.