Privacy Preserving Continuous Big Data Publishing
Salheddine Kabou, Laid Gasmi, Abdelbaset Kabou · 2024
The main challenge in big data analytics is preserving individual privacy. Anonymization models like k-anonymity and l-diversity are used to balance data privacy and utility during publishing. However, these models focus on a single data release and ensure a specific level of privacy. In practical big data applications, data publishing is more complex as it occurs continuously with new data collection, necessitating privacy for different releases. In this research, which is the primary research that focuses on the publication of continuous big data, a new distributed bottom-up approach is introduced to achieve the m-invariance privacy model in continuous big data scenarios. This approach is based on data insertion and splitting. Data records from various workers are inserted into an improved ascending R-tree generalization to minimize information loss. The second phase involves splitting overloaded nodes by reducing overlaps among resulting partitions according to the m-invariance model. Experimental results demonstrate significant improvements in data utility and execution time, compared to existing approaches in literature.