A combined random noise perturbation approach for multi level privacy preservation in data mining

Abstract

Now a Days huge volume of personal and sensitive data is collected and retrieved by various enterprises like social networking system, health networks, financial organizations and retailers. There are three main entities such as data owner; the database service provider and the client are mainly involved in this type of outsourced based data model. So that is more essential for the privacy preservation of the owner. Privacy preservation is a main challenging area in data mining. In that, Data based privacy perturbation technique is the standard model which performs the data transformation process before publishing the data. This paper proposes Additive Multiplicative Perturbation Privacy Preserving Data Mining (AM-PPDM) which is suitable for multiple trust level. In that, the random noise perturbation is applied to individual values before the data are published. This hybrid approach improves the privacy guarantee value during the reconstruction process. In AM-PPDM, the generated random Gaussian noise multiplied with the original data to produce different perturbed copies at various trust levels. By implementing this approach, the diversity attack is completely avoided during the reconstruction process.