Enhanced Applicability of Privacy Preservation for Perturbed Data in Multi-Partitioned Data Set

摘要

The perturbation technique has been widely considered for privacy preserving in data mining for different datasets. Generally, multi-partitioned datasets comprises of both vertical and horizontal data sets which is being a current demand of e-Business and e-Commerce data mining environment. In perturbation process, arbitrary noise from a recognized distribution is processed as privacy susceptible data, prior the data is thrown to the data miner. Consequently, the data miner rebuilds estimation to the unique data distribution from the perturbed data and exercises the renovated delivery for data mining principles. Owing to the count of noise, loss of information versus conservation of privacy is a constant transaction in the perturbation based techniques. The question is to what level the users are disposed to cooperate with their privacy? This is a preference that amends from individual to individual. To assess a tradeoff among data privacy and simplicity of individual's data, the first research is to describe the data perturbation technique with validation and authentication. Diverse individuals may have diverse approaches towards confidentiality, based on traditions and cultures. Unfortunately, the earlier perturbation based privacy preserving data mining techniques do not permit the individuals to decide their preferred privacy levels. This is a negative aspect as privacy is an individual choice. In this study, researchers propose an individually adaptable perturbation model which enables the individuals to choose their own privacy levels. The effectiveness of the proposed model lies is the enhancement of the Applicability of Privacy Preservation for Perturbed Data in Multi-partitioned datasets (APPDM) demonstrated by diverse experiments conducted on both synthetic and real-world data sets. Based on the experimental evaluation, researchers propose a simple, valuable and resourceful method to construct data mining models from perturbed data and enhance the process of privacy preservation.