A Wavelet‐Based Approach to Preserve Privacy for Classification Mining*
提出一种基于小波变换的数据转换方法，在隐藏私人数据的同时保留原始分类模式，实验表明Haar和Daub-4变换能有效保护实值数据的隐私和分类模式。
ABSTRACT Despite the commercial success of data mining, a major drawback has been acknowledged across academic, industry, and government sectors, namely, the issue of violating the privacy of individuals. We propose a data transformation method based on wavelets to disguise private data while preserving the original classification patterns. Wavelet transformations have been used extensively in signal processing for data reduction, multiresolution analysis, and removing noise from data. In our implementation, two commonly used wavelet transforms, the Haar and the Daub‐4 transforms, are tested for pattern and privacy preservation in classification mining tasks. Empirical results confirm that the Haar and the Daub‐4 transforms preserve the classification patterns and preserve the privacy for real valued data.