Privacy-preserving high-dimensional data publishing based on differential privacy guarantees model
Yuwen Wang, Bo Shen, Jun Feng Yang, Xu Han, Shanshan Li · IET conference proceedings. · 2025
The differential privacy mechanism limits the complete capture of relationships and reduces data availability when releasing high-dimensional data. To this end, we propose a privacy-preserving data query model based on maximum information coefficient and machine learning. The high-dimensional data is pre-processed by incorporating the overall maximum information coefficient, and the data with solid correlation is selected as the candidate dataset. The sensitive data within the candidate dataset is protected through a combination of k-anonymity and differential privacy classification. Subsequently, a prediction model that satisfies differential privacy protection requirements is trained based on private data to generate released data. This approach utilizes data preprocessing to optimize the publishing of high-dimensional data under differential privacy, effectively addressing the limited availability in published results caused by incomplete capture of inter-data relationships during dimensionality reduction. Experimental results demonstrate that the proposed model exhibits superior usability compared to similar methodologies.