ENHANCING A PERSONAL DATA ANONYMIZATION METHOD BY ATTRIBUTE VALUE PERMUTATION

Purpose of the study: analyze an adversary’s capabilities to maliciously exploit data anonymized by the shuffling (permutation) method, and to develop countermeasures to mitigate the potential misuse of such anonymized data. Methods of research: comparison and juxtaposition, scenario analysis, systems analysis, probability theory, algorithm theory, and experimental demonstration. Result(s): the article examines a method of personal data anonymization based on shuffling the values of attributes containing subject identifiers. It is demonstrated that the shuffling method preserves the analytical value of data while simultaneously reducing the risk of direct subject re-identification. However, vulnerabilities of the shuffling method were identified, associated with the possibility of adversarial exploitation of anonymized datasets. To improve security, a modification of the method is proposed, based on expanding the set of identifiers with artificially generated, syntactically relevant values. This approach reduces the likelihood of a successful attack and diminishes the market value of leaked datasets on data black markets. In addition, the study considers the use of a trusted intermediary (an anonymization center) that accumulates large identifier dictionaries and ensures the isolation of personal data processing, thereby enabling the application of the proposed modification of the shuffling method without requiring artificial generation of personal identifiers on the part of the data operator. Scientific novelty: the novelty of the study lies in identifying security threats to personal data subjects that arise when data are anonymized using the shuffling method, as well as in proposing a way to mitigate these threats by enhancing the shuffling method through the expansion of the set of shuffled values.