Scalable Distributed Data Anonymization for Large Datasets

Abstract

k-Anonymity and -diversity are two well-known privacy metrics that guarantee protection of the respondents of a dataset by obfuscating information that can disclose their identities and sensitive information. Existing solutions for enforcing them implicitly assume to operate in a centralized scenario, since they require complete visibility over the dataset to be anonymized, and can therefore have limited applicability in anonymizing large datasets. In this paper, we propose a solution that extends Mondrian (an efficient and effective approach designed for achieving k-anonymity) for enforcing both k-anonymity and -diversity over large datasets in a distributed manner, leveraging the parallel computation of multiple workers. Our approach efficiently distributes the computation among the workers, without requiring visibility over the dataset in its entirety. Our data partitioning limits the need for workers to exchange data, so that each worker can independently anonymize a portion of the dataset. We implemented our approach providing parallel execution on a dynamically chosen number of workers. The experimental evaluation shows that our solution provides scalability, while not affecting the quality of the resulting anonymization.

Achieving k-anonymity of a large-scale database in a distributed memory environment

“…The H-PGPkAA algorithm showed good acceleration with slightly equal information loss as that of the GkAA algorithm. Additionally, we believe that the data reorganization used in these algorithms can be applied to improve data utility while achieving k-anonymity in other distributed parallel algorithms employing horizontal data partitioning, including the GCCG algorithm [10] or the extended Mondrian algorithm [31].…”

“…There has been extensive research conducted on distributed privacy preserving data publishing over the last twenty years [27]- [31]. Within this domain, the primary challenges revolve around data partitioning and aggregation strategies.…”

“…In a later study conducted in 2022, researchers explored an approach to scalable data anonymization, as detailed in [31]. This approach was designed to address the challenges of anonymizing large datasets within a distributed memory environment, specifically utilizing Apache Hadoop.…”

“…This challenge becomes particularly pronounced as the number of processors (or workers) and the value of k (the anonymity constraint) increase, or when dealing with relatively small datasets. The earlier study conducted and detailed in [31] was restricted to a mere 12 workers. Consequently, this limitation failed to provide a comprehensive view of the information loss, particularly for higher k values, such as k = 100, resulting from excessive dataset partitioning.…”

“…It is important to acknowledge that the partitioning process may potentially introduce some disruption to the obtained results, with a slight increase in terms of information loss [10], [31] compared to the centralized GkAA algorithm [12]. However, we anticipate that the time savings achieved through our method will outweigh this trade-off.…”