Implementación y comparación de algoritmos de anonimización de datos en diversos contextos

Abstract

Abstract: Data is a valuable resource today and must be carefully handled to avoid its interception or misuse. For this reason, the need arises to implement techniques of anonymization that ensure the privacy of user data, eliminating or encrypting personal identifiers to prevent information leaks. This work analyzes and compares anonymization algorithms such as Mondrian, Incongnito and Datafly, which allow patterns in data to be identified through analysis techniques generalization or suppression, resulting in encryption or anonymization. The results show that Mondrian obtains the best regularity in the distribution of tuples, preserving the diversity of the anonymized data. Datafly approaches Mondrian's results, although with greater loss of information. Incognito, although more accurate and versatile has a higher computational cost, which limited its ability to anonymize a set of quasi-identifying attributes in comparison to Mondrian and Datafly.