Anonymizing high dimensional data by taxonomy free grouping

Privacy protection in publishing high dimensional data is a challenging problem. Surprisingly, there are very few works on this problem. Nevertheless, the latest approach proposed so far suffers two drawbacks, namely introduction of excessive information loss and dependence on a given generalization taxonomy. To address the issues, this paper proposes a taxonomy free grouping approach for anonymizing high dimensional data. This approach assigns transactions into groups based on the hamming distances among transactions, and derives a collection of item bags to represent each transaction group. Experiments on real world data show that this approach outperforms the state of the art approach.