Anonymizing location-based RFID data

In this paper, we study the problem of anonymizing high dimensional location-based RFID data for mining or research purposes. We consider the case where RFID cards are used for purchasing in place of magnetic cards. Databases containing such transactions of card holders could be very huge in number of records (equals to number of users) and dimensions (could be equal to the domain of locations where users are allowed to use their cards). This huge database containing user's purchasing history can be mined to find interesting knowledge. At the same time publication of data would cause re-identification attacks by adversaries who have partial knowledge about transactions. Therefore, before publishing transactional data, it should be made k-anonymous. However, traditional k-anonymity methods were designed to k-anonymize low dimensional databases and are not scalable much to produce good results when it comes to k-anonymous large high dimensional databases. In this paper, we provide a solution modeling k-anonymity principle to protect the privacy in publication of high dimensional databases. We propose greedy approach, which scales much better and in most cases finds solution close to the optimal. The proposed algorithm is experimentally evaluated.