Preventive Inference Control in Data-centric Business Models

Abstract

Inference control is a modern topic in data usage management, especially in the context of data-centric business models. However, it is generally not well understood how protection mechanisms could be designed to protect the users. The contributions of this paper are threefold: firstly, it describes the inference problem and relate it to protection mechanisms; secondly, it reports on a simple mechanism to provide a-priori inference protection; thirdly, it discusses on the drawbacks of such a mechanism, as well as on the acceptance it had on a preliminary, controlled field study. In particular, the study shows that, contrary to our expectations, participants prefer an a-posteriori approach based upon audits to detect whether inferences happened.