Data Perturbation Method Based on Contrast Mapping for Reversible Privacy-preserving Data Mining

Data mining has become an important service in cloud computing. Privacy-preserving schemes must be applied to the original data before data owners provide data publicly or send the data to remote servers for mining in order to avoid improper disclosure of privacy data. Previous studies have applied data perturbation approaches to modify the content of the original data; however, this might affect the accuracy of the mining results. To solve this issue, this study develops the reversible privacy contrast mapping (RPCM) algorithm, which applies the reversible data hiding techniques used in image processing to perturb and restore data. Furthermore, to identify whether perturbed data have been modified without authorization, RPCM allows users to embed watermarks in the data. The experimental results show that the knowledge contained in the data perturbed using RPCM is similar to that in the original data. The privacy disclosure risk does not increase when the degree of data perturbation increases.