Quantitative evaluation model of desensitization algorithm

Today, with the explosive growth of data volume, data has become an increasingly important strategic resource for various enterprises, but Big Data not only facilitates human society, but also brings huge privacy leakage problem. Because the data in business production often contains a large amount of user sensitive privacy information, in order to avoid the privacy leakage of these data in the test system and other environments, enterprises begin to pay more attention to desensitization technology. Because desensitization is mainly used in the process of software development in the business field, and is often used to generate internal test data, most of the results are classified or protected by patents. In addition, the evaluation of desensitization algorithm mostly stays at the level of general and qualitative evaluation. At present, there is no intuitive, accurate and universal quantitative evaluation standard. In this paper, aiming at the pain point of desensitization evaluation model, a data oriented quantitative desensitization evaluation model based on association rule mining method is designed. The model does not need to understand the desensitization algorithm design process, but only faces the data set before and after desensitization, which reduces the risk of desensitization algorithm leakage. In addition, the model also covers the common fields of various industries, and accurately and intuitively describes the availability of desensitization data in the subsequent test environment in the form of quantitative numbers, which provides an innovative idea for the unification of desensitization algorithm evaluation standards.