On Minimality Attack in Privacy Preserving Data Publishing

Abstract

"Data Publishing has become much concern in recent yerars for protecting the individual privacy. Information about the individuals is collected from various domain and is being published public. We can extract lot of wealth of information by collecting and sharing personal information. For example traditional organization like hospitals and census collect information from individuals and publish them. Census data provides us information which is used for demographic an economic research. Hospitals provide information to let us know how diseases spread and the diseases according to age, ender and so on. And nobody want to leak from which disease they are suffering from.For this reason, these organizations strive to publish the data such that it discloses as much statistical information as possible while preserving the privacy of individuals who contribute to the data. Data collection agencies publish information to facilitate research. The protection of individual privacy is much important while publishing the data. From the recent studies it shows that the adversary may possess a lot of extra knowledge called background knowledge about the individuals. The knowledge of the adversary and the algorithm used for protecting the privacy may lead to loss of much more information from the published table. In order to preserve privacy at the same time balancing the utility is a difficult task. Therefore, all the mechanisms try to minimise the level of nonymization thus becoming a reason to launch attacks and such kind of attack is called minimality attack.

In this thesis work, we devise an algorithm to provide a feasible solution against Minimality Attack. The algorithm is built on k-anonymity principle and l-diversity principle. The algorithm mainly concentrates on removing the attack despite the attack being present in many existing algorithms. We experiment our algorithm on medical data set which available on the public repository."