Towards Learning Privacy Policies

Abstract

With the proliferation of personal computing devices users are creating a variety of digitized personal information, from personal contact databases and multimedia content to context data such as location, activity and mood. Preventing unintended disclosure of such information is a key motivator for developing privacy management frameworks. It is equally critical that protecting privacy does not prevent users from completing essential tasks. Current efforts in privacy management have focussed on notations for privacy policy specification and on user interaction design for privacy management. However, little has been done to support automated analysis and learning of privacy policies. We advocate an approach based on inductive logic programming (ILP) for automatic learning of privacy policies. ILP is preferred over statistical learning techniques because it produces rules (privacy policies) which are comprehensible to the user and amenable to automated analysis.