Design and Application of the Digital Human Identity Attribute Data Synthesis System

In the era of big data, increasing demand for identity data sharing in fields like healthcare and tourism, along with stricter cybersecurity regulations, has made generating realistic digital human identity attribute data and ensuring its efficient use a critical challenge. However, existing methods struggle with complex data types and distributions, often neglecting the social and logical consistency of the synthetic data. To address this, we propose a Conditional Generative Adversarial Network (CGAN)-based system for synthesizing digital human identity attribute data. The system features three modules: data preprocessing, data synthesis, and data verification. The preprocessing module uses a composite encoder to handle discrete, continuous, and mixed data types, with a Variational Gaussian Mixture Model (VGM)-based normalization for continuous variables. The synthesis module introduces conditional vector-based sampling to resolve data imbalance while preserving data distribution characteristics, enhancing the realism of the generated data. The verification module uses association rule mining for secondary validation, ensuring logical consistency. This modular design improves scalability, paving the way for multimodal data synthesis. Experimental results show that the system excels in statistical similarity and machine learning utility, with the verification module boosting rule retention by 15%, significantly improving data quality and practical value. This study provides an effective solution with important theoretical and practical implications for digital human identity attribute data synthesis.