Anonymizing Sequential Releases under Arbitrary Updates
Résumé
In today's global information society, governments, compa- nies, public and private institutions and even individuals have to cope with growing demands for personal data pub- lication from scientists, statisticians, journalists and many other data consumers. Current researches on privacy-preserving data publishing by sanitization focus on static dataset, which have no updates. In real life however, data sources are dy- namic and usually the updates in these datasets are mainly arbitrary. Then, applying any popular static privacy-preserving technique inevitably yields to information disclosure. Among the few works in the literature that relate to the serial data publication, none of them focuses on arbitrary updates, i.e. with any consistent insert/update/delete sequence, and es- pecially in the presence of auxiliary knowledge that tracks updates of individuals. In this communication, we first high- light the invalidation of existing algorithms and present an extension of the m-invariance generalization model coined τ-safety. Then we formally state the problem of privacy- preserving dataset publication of sequential releases in the presence of arbitrary updates and chainability-based back- ground knowledge. We also propose an approximate algo- rithm, and we show that our approach to τ-safety, not only prevents from any privacy breach but also achieve a high utility of the anonymous releases. pdf on : http://www.edbt.org/Proceedings/2013-Genova/papers/workshops/a21-anjum.pdf