@inproceedings{oai:ipsj.ixsq.nii.ac.jp:00077972,
 author = {五十嵐, 大 and 千田, 浩司 and 高橋, 克巳 and Dai, Ikarashi and Koji, Chida and Katsumi, Takahashi},
 book = {コンピュータセキュリティシンポジウム2011 論文集},
 issue = {3},
 month = {Oct},
 note = {近年, プライバシーを保護しつつデータを活用する研究が活発に行われており, その中でもk-匿名性を実現するk-匿名法は特に注目されている. 一方確定的アルゴリズムを用いるk-匿名化でなく, データのランダム化による匿名化手法も研究されている.  CSS2009において筆者らはランダム化したデータベース上のk-匿名性, Pk-匿名性を定義しその実現方法を示した. しかし対象は例えば性別のように値の種類の少ない, カテゴリ属性のみであった. 本稿ではこれを身長や年収などの数値属性に拡張する. 具体的には, ラプラスノイズと呼ばれるノイズの加算がPk-匿名性を満たすことを示し, 有用性とPk-匿名性の両立の観点からは本ノイズが最適であることを示す., Recently, studies on the utilizing data privately, are widely conducted. Among them, kanonymity is especially paid attention to. On the other hand, in community of database and statistics, not only k-anonymity, which adopt non-probabilistic algorithm, but also anonymization using data randomization is expected and has been studied. In CSS2009, we defined Pk-anonymity, which is k-anonymity on a randomized database, and showed a method for satisfying Pk-anonymity. However, it was applicable only to categorical attributes. In this paper, we extend it to numerical attributes. First, we show that the addition of a noise on a unbounded numerical attribute can not satisfy Pkanonymity, and that even on a bounded attribute, the addition of the uniform distribution and the normal distribution proposed by Aggrawal et al. can not satisfy it. Finally, we show that a noise called Laplase distribution satisfies Pk-anonymity on a bounded numerical attribute.},
 pages = {450--455},
 publisher = {情報処理学会},
 title = {数値属性における, k-匿名性を満たすランダム化手法},
 volume = {2011},
 year = {2011}
}