Synergizing Privacy and Utility in Data Analytics Through Advanced Information Theorization

TL;DR

This paper introduces advanced information-theoretic algorithms for privacy-preserving data analytics, balancing data utility with privacy across high-dimensional and structured datasets, and setting new benchmarks in the field.

Contribution

It presents a novel framework with three algorithms—noise infusion, VAE, and EM—for enhanced privacy and utility in diverse data types, advancing current methods.

Findings

Significantly reduces mutual information between sensitive attributes and data.

Achieves high privacy protection while maintaining data utility.

Sets new benchmarks for privacy-utility trade-offs in data analytics.

Abstract

This study develops a novel framework for privacy-preserving data analytics, addressing the critical challenge of balancing data utility with privacy concerns. We introduce three sophisticated algorithms: a Noise-Infusion Technique tailored for high-dimensional image data, a Variational Autoencoder (VAE) for robust feature extraction while masking sensitive attributes and an Expectation Maximization (EM) approach optimized for structured data privacy. Applied to datasets such as Modified MNIST and CelebrityA, our methods significantly reduce mutual information between sensitive attributes and transformed data, thereby enhancing privacy. Our experimental results confirm that these approaches achieve superior privacy protection and retain high utility, making them viable for practical applications where both aspects are crucial. The research contributes to the field by providing a flexible and effective strategy for deploying privacy-preserving algorithms across various data types and establishing new benchmarks for utility and confidentiality in data analytics.