Towards Tracing of k-Anonymized Datasets

In this paper, we consider a simple procedure for finding the source of a leaked dataset and evaluate its applicability to general k-anonymization datasets. The focus of the paper is to identify each anonymized dataset, and thus how to generate a large number of different datasets from the original table is an important issue. We evaluate the number of possible anonymized datasets and consider how to increase the number of anonymized datasets by performing a modification process before anonymization. We also consider a new security notion and security evaluation against several types of attacks. Through our experiments, we confirmed that we could generate some k-anonymized datasets that satisfy a certain level of security against the attacks and even a simple modification process was shown to be an effective way of increasing the number of possible anonymized datasets. This paper is the first step to realizing traceability of a leaked k-anonymized dataset.