Two-Fold Differentially Private Mechanism for Big Data Analysis

Differential privacy (DP) has emerged as a gold standard for privacy preservation in many applications, particularly in light of recent advancements in machine learning for big data for cloud services and the growing threat of privacy attacks. However, the addition of random noise to data for privacy preservation often results in decreased data quality and utility. To address this challenge, we propose a novel twofold differentially private data generation method that leverages the power of denoising autoencoders to preserve higher data quality and utility. Our approach combines traditional additive differential privacy with a novel reductive differential privacy approach that uses a denoising autoencoder to restore the original distribution of the data, increasing the data utility in machine learning tasks. We also experimentally show the effectiveness of the proposed method by experimental evaluation.