ACCUMULATION OF DATA PERTURBATION TECHNIQUES FOR PRIVACY PRESERVING DATA CLASSIFICATION

In this work propose different data perturbation techniques that deal with multi dimensional data perturbation and achieve high privacy guarantee and zero- loss of accuracy, various perturbation techniques are used for different classifiers. The main aim is to perturb the multi dimensional data to preserve multi dimensional information. There are three different data perturbation techniques are proposed and their performance are compared. The three data perturbation techniques are Additive Perturbation, Multiplicative Perturbation, Rotation Perturbation. The Proposed system shows that the multiplicative and rotation perturbation makes balance between the privacy and accuracy is done effectively. The performance over the perturbed dataset is greater than the original dataset. Using the classifiers C4.5 and SVM the privacy and accuracy of the proposed techniques are compared.