Learning Privacy Preferences

Abstract

This paper suggests a machine learning approach to preference generation in the context of privacy agents. With this solution, users are relieved from the complex task of specifying their preferences beforehand, disconnected from actual situations. Instead, historical privacy decisions are used as a basis for providing privacy recommendations to users in new situations. The solution also takes into account the reasons why users act as they do, and allows users to benefit from information on the privacy trade-offs made by others.