An Efficient Hybrid Data Perturbation Approach to Preserve Privacy

Sharing of electronic data among organizations may disclose individual's sensitive information. Privacy preserving data mining (PPDM) plays a vital role in sensitive data protection. We propose an efficient hybrid data perturbation approach to address the privacy of both numeric and categorical attributes in electronic records. The categorical data are converted to numeric then original data are perturbed using S-based fuzzy membership function. Then attribute based multiplicative data perturbation is performed on different real world data sets to generate distorted data. Our proposed approach has proved its efficiency with K-means clustering and minimum distance classification of data before and after data transformation.