Enforcing Confidentiality and Visibility Constraints

The most straightforward understanding of, and the first requirement for, protecting privacy when releasing a data collection is indeed the protection of the sensitive data included in the release. However, privacy protection should not prevent recipients from performing legitimate analysis on the released dataset, and should ensure adequate visibility over non sensitive information. In this chapter, we illustrate a solution allowing a data owner to publicly release a dataset while satisfying confidentiality and visibility constraints over the data, expressing requirements for information protection and release, respectively, by releasing vertical views (fragments) over the original dataset. We translate the problem of computing a fragmentation composed of the minimum number of fragments into the problem of computing a maximum weighted clique over a fragmentation graph. The fragmentation graph models fragments, efficiently computed using Ordered Binary Decision Diagrams (OBDDs), which satisfy all the confidentiality constraints and a subset of the visibility constraints defined in the system. To further enrich the utility of the released fragments, our solution complements them with loose associations (i.e., a sanitized form of the sensitive associations broken by fragmentation), specifically extended to safely operate on multiple fragments. We define an exact and a heuristic algorithm for computing a minimal and a locally minimal fragmentation, respectively, and a heuristic algorithm to efficiently compute a safe loose association among multiple fragments. We also prove the effectiveness of our proposals by means of extensive experimental evaluations.