Differential Privacy in Privacy-Preserving Big Data and Learning: Challenge and Opportunity

TL;DR

This paper reviews the current state of differential privacy in big data and machine learning, highlighting its strengths, limitations, and open challenges, and suggests new directions for integrating it with other privacy-preserving techniques.

Contribution

It identifies key limitations and challenges of differential privacy in various applications and proposes combining it with dimension reduction and secure multiparty computing for improved privacy models.

Findings

Differential privacy is widely adopted but has limitations in certain scenarios.

There are misunderstandings and misuse of DP in specific applications.

Combining DP with other techniques offers promising new privacy models.

Abstract

Differential privacy (DP) has become the de facto standard of privacy preservation due to its strong protection and sound mathematical foundation, which is widely adopted in different applications such as big data analysis, graph data process, machine learning, deep learning, and federated learning. Although DP has become an active and influential area, it is not the best remedy for all privacy problems in different scenarios. Moreover, there are also some misunderstanding, misuse, and great challenges of DP in specific applications. In this paper, we point out a series of limits and open challenges of corresponding research areas. Besides, we offer potentially new insights and avenues on combining differential privacy with other effective dimension reduction techniques and secure multiparty computing to clearly define various privacy models.