Hardware Enforced Statistical Privacy

摘要

The Internet of Things will result in users generating vast quantities of data, some of it sensitive. Results from the statistical analysis of sensitive data across wide ranges of demographics will become ever more useful to data analysts and their clients. The competing needs of the two groups—data generators with their desire for privacy and analysts with their desire for inferred statistics—will be met through the use of statistical privacy techniques. The question, then, is how can we ensure that the statistical methods are applied in a trustable manner? In this paper we discuss some of the complications and consequences of ensuring both trust and privacy through the immutability of hardware, providing a desiderata for a hardware privacy platform.