DESIGN AND IMPLEMENTATION OF A DATA ANONYMIZATION ALGORITHM FOR PRIVACY PRESERVATION WITH ENHANCED DATA USABILITY
Dhananjay M. Kanade, Snehal M. Kamalapur, Shirish S. Sane
Abstract: The rapid growth of digital data in healthcare, transportation, finance, and social sciences has increased concerns about re-identification and unauthorized inference. This study designs and evaluates a synthetic-data framework using decision-tree, GAN, and CopulaGAN generators on the Adult, ATUS, and FARS datasets. Evaluation focuses on classification utility and structural fidelity. Decision-tree synthetic data achieve the highest synthetic accuracy on Adult and ATUS, whereas CopulaGAN achieves the highest synthetic accuracy on FARS. Privacy risks, including re-identification, disclosure, membership inference, attribute inference, and record linkage, were not quantitatively assessed. Therefore, the findings support utility retention but do not demonstrate privacy improvement or an optimal privacy-utility trade-off. © 202x Published by Faculty of Engineering
Keywords: Data anonymization; Synthetic data; Privacy preservation; Machine learning; GAN; CopulaGAN; Decision tree; Data usability
DOI: 10.24874/PES08.03.030
Recieved: 06.05.2026 Revised: 29.06.2026 Accepted: 03.08.2026
UDC:
Reads: 3 
