Privacy Preserving Classification using 4-Dimensional Rotation Transformation

This paper focuses on privacy-preserving of the data while extracting some meaningful information from the large lumps of data available. Many techniques have been used for privacy-preserving of the data while mining but to balance it with data utility has always been a difficult task. In this paper four-dimensional rotational transformation to transform the numeric data into disguised format to preserve its privacy by hiding its sensitivity has been used. The non-numeric data could be preserved using cryptographic algorithms. The data before transformation is normalized using min-max normalization and Geometric Data Perturbation method, which fractionates the numeric data into 2 groups and rotates these groups simultaneously along two planes vis-a-vis xy-plane and zw-plane using four-dimensional rotational matrices. To rigidify the security level, the angle whose variance value is high is selected. By rigidifying the security level it becomes difficult to extort the actual data from the transformed data format. In the proposed method the privacy and utility of data are preserved using 4D Rotation Transformation.