Date of Award
Doctor of Philosophy (PhD)
College of Computing and Engineering
The data mining sanitization process involves converting the data by masking the sensitive data and then releasing it to public domain. During the sanitization process, side effects such as hiding failure, missing cost and artificial cost of the data were observed. Privacy Preserving Data Mining (PPDM) algorithms were developed for the sanitization process to overcome information loss and yet maintain data integrity. While these PPDM algorithms did provide benefits for privacy preservation, they also made sure to solve the side effects that occurred during the sanitization process. Many PPDM algorithms were developed to reduce these side effects. There are several PPDM algorithms created based on different PPDM techniques. However, previous studies have not explored or justified why non-traditional side effects were not given much importance.
This study reported the findings of the side effects for the PPDM algorithms in a newly created web repository. The research methodology adopted for this study was Design Science Research (DSR). This research was conducted in four phases, which were as follows. The first phase addressed the characteristics, similarities, differences, and relationships of existing side effects. The next phase found the characteristics of non-traditional side effects. The third phase used the Privacy Preservation and Security Framework (PPSF) tool to test if non-traditional side effects occur in PPDM algorithms. This phase also attempted to find additional unknown side effects which have not been found in prior studies. PPDM algorithms considered were Greedy, POS2DT, SIF_IDF, cpGA2DT, pGA2DT, sGA2DT. PPDM techniques associated were anonymization, perturbation, randomization, condensation, heuristic, reconstruction, and cryptography. The final phase involved creating a new online web repository to report all the side effects found for the PPDM algorithms. A Web repository was created using full stack web development. AngularJS, Spring, Spring Boot and Hibernate frameworks were used to build the web application. The results of the study implied various PPDM algorithms and their side effects. Additionally, the relationship and impact that hiding failure, missing cost, and artificial cost have on each other was also understood. Interestingly, the side effects and their relationship with the type of data (sensitive or non-sensitive or new) was observed. As the web repository acts as a quick reference domain for PPDM algorithms. Developing, improving, inventing, and reporting PPDM algorithms is necessary. This study will influence researchers or organizations to report, use, reuse, or develop better PPDM algorithms.
Hima Bindu Sadashiva Reddy. 2022. Exploring the Existing and Unknown Side Effects of Privacy Preserving Data Mining Algorithms. Doctoral dissertation. Nova Southeastern University. Retrieved from NSUWorks, College of Computing and Engineering. (1179)