Privacy Protection in Data Mining

Data mining as one of the important means to discover interesting and potential useful patterns or knowledge from large data sources has been widely used for improving business intelligence. Since some data items may be specific to individuals, companies increasingly pay attention to privacy issues while implementing business intelligence solutions. In this paper, we present a framework for privacy-enhancing data mining and develop such privacy-enhancing technologies as attribute selection, discretization, fixed-data perturbation, probability distribution, and randomization. Specifically, we address the issue of privacy protection through using the attribute selection, discretization, and randomization techniques and give an example of inducing the decisiontrees from training data in which the values of sensitive attributes have been modified by using these techniques. The results show that we can achieve comparative predictive accuracies without accessing the actual values of the sensitive attributes.