Um Estudo Comparativo entre Algoritmos de Proteção da Privacidade e Segurança de Dados Aplicado à Bases de Dados na Área de Saúde

Abstract

The development of new technologies and the growing increase in the volume of data collected and stored in databases in the health area, together with the need to share this data created benefits and oportunities to the knowledge based decision making process, pushing the application of data mining techniques to extract usefull information. Despite the benefits, sharing this data in its original form could jeoperdize the privacy of the records. Therefore, some institutions that research in the health care area and have the necessity to use various data sources, could hesitate to share their data amongst them. In an attempt to solve this problem, several techniques focused on data privacy protection on data mining have been proposed in the literacture, in which the most diverse methods are applied. Due the importance of the application of these techniques, in this paper a comparative analisys between anonymization, perturbation and DRBT techniques is realized, in search to find which technique is more efficient to preserve privacy of data submitted to cluster analisys. The obtained results display that the evaluated techniques are effective in preserving data privacy, showing that the greater the privacy obtained, the lower the accuracy of clusters. Among the analyzed and tested techniques, the perturbation technique proved to be more effective, since it maintains a balance between the data privacy and the accuracy of data mining results.