Several formal languages have been proposed to encode privacy policies, ranging from the Platform for Privacy Preferences (P3P), intended for communicating privacy policies to consumers over the web, to the Enterprise Privacy Authorization Language (EPAL), intended to enable policy enforcement within an enterprise. However, current technology does not allow an enterprise to determine whether its detailed, internal enforcement policy meets its published privacy promises. We present a data-centric, unified model for privacy, equipped with a modal logic for reasoning about permission inheritance across data hierarchies. We use this model to critique two privacy preference languages (APPEL and XPref), to justify P3P's policy summarization algorithm, and to connect privacy policy languages, such as EPAL. Specifically, we characterize when one policy enforces another and provide an algorithm for generating the most specific privacy promises, at a given level of detail, guaranteed by a more detailed enforcement policy.
展开▼
