High Performance Privacy Protection Hardware

As more and more person-specific data like health information becomes available, increasing attention is being paid to confidentiality and privacy protection. One proposed measure of confidentiality is k-anonymity, where a dataset is k-anonymous if each record is identical to at least (k-1) others in the dataset. All known algorithms for k-anonymization or even checking the degree of k-anonymity of a dataset are too expensive in both running time and memory accesses required for terabyte-sized datasets. In this paper, we present a novel solution that uses both algorithm parallelism and reconfigurable hardware to achieve acceptable performance on both counts.