Privacy-aware data publishing against sparse estimation attack

Recently a number of privacy-preserving data publishing techniques have been proposed to protect the privacy of released data. In this paper, we study how to protect unreleased data privacy during data publishing. Our experiment results show that the unreleased data could be well estimated when an attacker leverages sparse estimation techniques, even when a large amount of noise is randomly added to the released data. To address unreleased data privacy while guaranteeing the utility of released data, we propose a privacy-aware structural data publishing framework against sparse estimation attack. Specifically, we present a nonzero element Gaussian random noise addition strategy, which is realized by maximizing global information loss between original data and noisy data. Furthermore, we deduce the upper bound of the number of released data that could be published, which acts as the criterion to guarantee the unreleased data privacy. Our experiment results show that the proposed framework is able to protect unreleased data privacy with desirable performance.