Efficient Privacy Preservation of Big Data Using Random Number Generators and Geometric Data Transformations

Abstract

Recent trends indicate that the volume of data stored in repositories is exploding due to the development of technology and the pervasive usage of web-based activities. This vast collection of data may contain personal information, which may occasionally pose privacy problems. The purpose of this research work is to establish privacy-preserving data publication mechanisms that are applicable to any numerical attributes and to design a distributed data model to process big data. Four distinct classifiers, namely Decision Tree, Naive Bayes, Adaboost, and KNN, are used to evaluate the classification accuracy of the suggested model. A geometric data perturbation-based method (RSUGP) and random number generators is used to protect sensitive data. For geometric data perturbation, a noise model based on random number generators was utilized instead of random noise or Gaussian noise. A Graphical neural network (GNN) is utilized for training, testing, and classification with an accuracy of 93%. Experiments indicate that the proposed strategy is superior to the other three in terms of attack resistance, classification precision, and runtime.