| dc.description.abstract | Nowadays, more than ever data comes from different sources, providing an opportunity against various threats. Different privacy issues such as linkage, data breaches, false identities, and other frauds concern both people and organizations. In order to deal with such a problem, the term Privacy-preserving approach with Differential Privacy as the leading mechanism was invented. Local Differential Privacy can be achieved by adding randomized noise into the dataset, however, too much noise could affect the data quality and value of the dataset. The thesis aims to introduce a visualization system that can help users understand how the privacy mechanism affects data and adjust the noise added by those algorithms. In order to provide such a framework, it was decided to implement a visualization system that uses an alternative and simple inspiration of the local differential privacy mechanism to provide visual analysis on specific data. Moreover, specific metrics for inspecting data privacy and utility will be used to evaluate the mechanism's performance. By creating an interactive visualization system that offers to adjust the epsilon parameter with slider and instantly presenting different graphics, users will understand how privacy affects data utility and the opposite. The thesis combines different layers of evaluation that comprehend experts, case study, and technical evaluations to validate the project. As a result, the solution was recognized as a visual analytics approach to explaining the effect of noise-injection levels on a specific dataset by taking four layers of evaluation methods. In addition, the experts agreed that the project contributes to defining privacy-preserving visual analytics as the first approach that explains the means of data privacy and utility tradeoff with the visualization system. | |
