The Impact of Data Perturbation Techniques on Data Mining

Data perturbation is a data security technique that adds 'noise' to databases to allow individual record confidentiality. This technique allows users to ascertain key summary information about the data while preventing a security breach. Four bias types have been proposed which assess the effectiveness of such a technique. However, these biases deal with simple aggregate concepts (averages, etc.) found in the database. In e-commerce applications, organizations are interested in applying data mining approaches to databases to discover additional knowledge about customers. In our study, we propose a fifth type of bias that may be added by perturbation techniques (Data mining Bias), and empirically test for its existence. Our results find support for this bias, and propose future research avenues that are appropriate for the emerging interdisciplinary field in data security, e-commerce and data mining.