A Multidimensional Evaluation Scheme for Data Desensitization Algorithm Capabilities
Ruihan Liu, Hui Zhu, Fengwei Wang, Hui Li, Fenghua Li · 2025
To address the challenges in evaluating data desensitization performance, this study proposes a multidimensional evaluation scheme for systematically quantifying desensitization algorithm capabilities. The scheme enables a multi-dimensional evaluation, yielding a comprehensive score and grade for the desensitization capability. Starting from four aspects of algorithm capabilities—reversibility, information deviation, information loss, and complexity—the scheme quantifies and grades the index values by applying benchmark desensitization algorithms to sampled datasets, thereby establishing an evaluation index system. Based on this system, along with the data modalities that desensitization algorithms can process, the categories of desensitization algorithms, and their application scenarios, a multidimensional evaluation scheme is proposed to achieve a comprehensive evaluation of desensitization algorithm capabilities. The feasibility of the evaluation scheme is validated by evaluating the capabilities of text replacement and numerical shift algorithms using a synthetic personal information dataset.