Privacy Preservation in Data Sharing: A Study on Differential Privacy

In the age of Big Data, the tension between data utility and individual privacy rights has become increasingly apparent. Nowadays, data is ubiquitous, while data storage is both accessible and affordable. This allows numerous entities to leverage this data for research purposes, machine learning models training, and statistics generation.

However, a critical question arises: how can this be accomplished while ensuring the protection of individuals’ privacy? This thesis examines the role of Differential Privacy as a promising approach to address these concerns, offering insights into its practical implementation and effectiveness in preserving privacy during data sharing processes.

The opening chapter of this thesis introduces the issue of online privacy, featuring pertinent real-life instances of breaches across various fields in recent years. Next, an examination of privacy-preserving methods utilized in earlier contexts will be undertaken.

In the second chapter, the definition and properties of Differential Privacy will be presented.

In the third chapter, various differentially-private mechanisms will be discussed.

In the fourth chapter, emphasis will be placed on executing experiments related to Differential Privacy, with the aim of discerning its applicable domains. Several frameworks with Differential Privacy guarantee will be developed, while the results will be evaluated in terms of the level of privacy achieved.

The fifth chapter will explore some real-world uses of Differential Privacy, along with the privacy budget parameters used by each.

The final chapter will provide a summary of the conducted work, extracting key points and presenting the future challenges of Differential Privacy.