Attribute-oriented Granulation for Privacy Protection

How to achieve a balance between data publication and privacy protection has been an important issue in infor- mation security for some years. When microdata is released to users, attributes that clearly identify individuals are generally removed. However, it is still possible to link released data with some public or easy-to-access database to obtain confidential information. Numerous techniques, such as generalization, sup- pression, and microaggregation, have been proposed to modify the released data to safeguard privacy. In this paper, we propose attribute-oriented granulation as a data modification method. We address the computational issue of searching for the most specific granulation that satisfies confidentiality requirements. A breadth-first search algorithm with basic pruning strategies is presented and its properties are investigated. The properties can be used to improve the efficiency of our algorithm. We also define the quantitative measures of data quality and safety, and apply evolutionary computation techniques to find the optimal granulation for privacy protection.