@techreport{oai:ipsj.ixsq.nii.ac.jp:00234043,
 author = {Akito, Yamamoto and Tetsuo, Shibuya and Akito, Yamamoto and Tetsuo, Shibuya},
 issue = {4},
 month = {May},
 note = {With the increasing amount of data in society, privacy concerns in data sharing have become widely recognized. Particularly, it is essential to protect personal attribute information. Although various differentially private methods based on randomized response have been proposed, there is a lack of studies on the mechanism for sharing multi-attribute data. The existing randomized response for this purpose uses the Kronecker product to perturb each attribute information in turn according to the respective privacy level but achieves only a weak privacy level for the entire dataset. Therefore, we first describe a privacy-optimized randomized response that guarantees the strongest privacy in sharing multi-attribute data. Thereafter, we present an efficient heuristic algorithm for constructing a near-optimal mechanism. The time complexity of our algorithm is O(k2), where k is the number of attributes. The experimental results demonstrate the utility of our methods., With the increasing amount of data in society, privacy concerns in data sharing have become widely recognized. Particularly, it is essential to protect personal attribute information. Although various differentially private methods based on randomized response have been proposed, there is a lack of studies on the mechanism for sharing multi-attribute data. The existing randomized response for this purpose uses the Kronecker product to perturb each attribute information in turn according to the respective privacy level but achieves only a weak privacy level for the entire dataset. Therefore, we first describe a privacy-optimized randomized response that guarantees the strongest privacy in sharing multi-attribute data. Thereafter, we present an efficient heuristic algorithm for constructing a near-optimal mechanism. The time complexity of our algorithm is O(k2), where k is the number of attributes. The experimental results demonstrate the utility of our methods.},
 title = {Efficient Heuristic Algorithm for Privacy-Optimized Randomized Response},
 year = {2024}
}