Privacy Preservation Using Multi-Context Systems and Default Logic J¨ urgen Dix 1⋆ , Wolfgang Faber 2⋆⋆ , and V.S. Subrahmanian 3 1 Department of Informatics Clausthal University of Technology 38678 Clausthal, Germany dix@tu-clausthal.de 2 Department of Mathematics University of Calabria 87030 Rende (CS), Italy wf@wfaber.com 3 Department of Computer Science University of Maryland College Park, MD 20742 vs@cs.umd.edu Abstract. Preserving the privacy of sensitive data is one of the major challenges the information society has to face. Traditional approaches focused on infrastruc- tures for identifying data which is to be kept private and for managing access rights to these data. However, although these efforts are useful, they do not ad- dress an important aspect: While the sensitive data itself can be protected nicely using these mechanisms, related data, which is deemed insensitive per se, may be used to infer sensitive data. This inference can be achieved by combining in- sensitive data or by exploiting specific background knowledge of the domain of discourse. In this paper, we present a general formalization of this problem and two particular instantiations of it. The first supports query answering by means of multi-context systems and hybrid knowledge bases, while the second allows for query answering by using default logic. 1 Introduction With the advent of the Internet and easy access to huge amounts of data, keeping sensi- tive data private has become a priority for distributed information systems. An example area in which privacy is at stake are medical information systems. Most databases have privacy mechanisms which are comparatively simple. Often, this boils down to keeping certain columns of the database hidden from certain types ⋆ The first author acknowledges that this work was partly funded by the NTH School for IT Ecosystems (NTH (Nieders¨ achsische Technische Hochschule) is a joint university consisting of Technische Universit¨ at Braunschweig, Technische Universit¨ at Clausthal, and Leibniz Uni- versit¨ at Hannover). ⋆⋆ Supported by M.I.U.R. within the PRIN project LoDeN. Some of the authors of this paper were funded in part by ARO grants W911NF0910206 and W911NF0910525 and W911NF1110344.