Quantitative Metrics for Balancing Privacy and Utility in Pseudonymized Big Data

The increasing demand for data utilization has renewed attention to the trade-off between privacy protection and data utility, particularly concerning pseudonymized datasets. Traditional methods for evaluating re-identification risk and utility often rely on fragmented and incompatible metrics, complicating the assessment of the overall effectiveness of pseudonymization strategies. This study proposes a novel quantitative metric—Relative Utility–Threat (RUT)—which enables the integrated evaluation of safety (privacy) and utility in pseudonymized data. Our method transforms various risk and utility metrics into a unified probabilistic scale (0–1), facilitating standardized and interpretable comparisons. Through scenario-based analyses using synthetic datasets that reflect different data distributions (balanced, skewed, and sparse), we demonstrate how variations in pseudonymization intensity influence both privacy and utility. The results indicate that certain data characteristics significantly affect the balance between protection and usability. This approach relies on simple, lightweight computations—scanning the data once, grouping similar records, and comparing their distributions. Because these operations naturally parallelize in distributed environments such as Spark, the proposed framework can efficiently scale to large pseudonymized datasets.