Towards Information Privacy for the Internet of Things

TL;DR

This paper proposes a nonparametric learning framework for privacy-preserving hypothesis testing in IoT networks, ensuring high privacy for private hypotheses while maintaining accurate public hypothesis inference.

Contribution

It introduces an empirical normalized risk concept and iterative algorithms for designing local privacy mappings without assuming joint distribution knowledge.

Findings

Achieves low error in public hypothesis detection

Ensures high error in private hypothesis detection

Provides theoretical guarantees with large training data

Abstract

In an Internet of Things network, multiple sensors send information to a fusion center for it to infer a public hypothesis of interest. However, the same sensor information may be used by the fusion center to make inferences of a private nature that the sensors wish to protect. To model this, we adopt a decentralized hypothesis testing framework with binary public and private hypotheses. Each sensor makes a private observation and utilizes a local sensor decision rule or privacy mapping to summarize that observation independently of the other sensors. The local decision made by a sensor is then sent to the fusion center. Without assuming knowledge of the joint distribution of the sensor observations and hypotheses, we adopt a nonparametric learning approach to design local privacy mappings. We introduce the concept of an empirical normalized risk, which provides a theoretical guarantee for the network to achieve information privacy for the private hypothesis with high probability when the number of training samples is large. We develop iterative optimization algorithms to determine an appropriate privacy threshold and the best sensor privacy mappings, and show that they converge. Finally, we extend our approach to the case of a private multiple hypothesis. Numerical results on both synthetic and real data sets suggest that our proposed approach yields low error rates for inferring the public hypothesis, but high error rates for detecting the private hypothesis.