A Differential Privacy Perturbation with Random Forest Classifier in Medical Database

In the era of information and technology, privacy become a major concern is in the area of collaborative mining. In the proposed method, the data are perturbed using differential private method to achieve confidentiality of data. Then, the perturbed data are classified using Random Forest classifier. The accuracy of the classifier is adjusted with respect to the threshold level to get the high throughput in utility mining. Meanwhile, the algorithm is compared with other ensemble classifier AdaBoost and Gradient boost classifier (GBC) for classification. In this paper, differential privacy and Gaussian noise privacy level are compared and analyzed to get the effects of both the perturbation technique on the private data and found that Random Forest classifier with differentially perturbed data shown a significant privacy and confidentiality without compromising its utility. Random Forest classifier, an adversary model classifies the quasi-identifier.