Data desensitization is the process of performing some special transformation on sensitive data to ensure data security. Data synchronization technology is widely used in relational databases to ensure data consistency and synchronization between different databases. The current data synchronization system is unable to preserve the original format of data while desensitizing sensitive data and is unable to adopt different desensitization methods for different types of sensitive data. In order to solve the above problems, a desensitization algorithm MixFF3 that preserves the original format of the data is proposed, which combines the data mixing and FF3-1 algorithms to achieve data desensitization. A data synchronization desensitization system is designed, which uses CDC technology to capture changed data in real time and can flexibly desensitize different kinds of data with multiple methods and synchronize them to the target database. We tested the performance of the MixFF3-based synchronous desensitization system for synchronous desensitization on data tables of different magnitudes. The results show that the system preserves the original format of the data and does not impose excessive overhead on the data synchronization task after desensitization using MixFF3 for a 100,000-record synchronization task.