Master's Thesis Marcus Land

Supporting diverse stakeholders to make informed decisions about the use of differential privacy with a web-based e-learning application

Abstract

Privacy has long been overshadowed by security but is becoming increasingly important as ever more data is collected, shared and analyzed. This has also been noticed by legislators, leading to new privacy laws like the GDPR. It has become evident that cybersecurity measures such as authentication and encryption are insufficient to ensure privacy. There is a need for dedicated privacy-preserving methods. However, recent research has shown that traditional approaches to privacy are deeply flawed, as they lack provable guarantees. This is addressed by differential privacy which emerged in 2006. Although it is regarded as a gold standard in privacy by research, a large part of the industry has not adapted it yet. The only exception being key players like Google, Microsoft or the US Census Bureau. Other and especially smaller companies seem to be hesitant in adopting it.

In this thesis, we want to investigate the reasons behind this imbalance. We evaluate the challenges and benefits and overall maturity of differential privacy to close the gap between research and practice. Even tough there are many tools available to facilitate the use of differential privacy, only few address the question whether differential privacy should be used in the first place. We want to guide practitioners in these earlier stages of their decision-making process. To do so, we look into the current state of differential privacy in the industry by studying the grey literature and conducting expert interviews. Based on these findings, we design learning material to support practitioners in making informed decisions about the use of differential privacy.