{"updated":"2025-01-19T18:51:41.661896+00:00","links":{},"id":208454,"created":"2025-01-19T01:09:50.832074+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00208454","sets":["6164:6165:6462:10428"]},"path":["10428"],"owner":"44499","recid":"208454","title":["差分プライバシと秘密計算の融合による秘匿性がデータ提供者の数に依存しない秘匿協調学習"],"pubdate":{"attribute_name":"公開日","attribute_value":"2020-10-19"},"_buckets":{"deposit":"b6bd4aef-22bf-46d0-9978-32581d0a96ae"},"_deposit":{"id":"208454","pid":{"type":"depid","value":"208454","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"差分プライバシと秘密計算の融合による秘匿性がデータ提供者の数に依存しない秘匿協調学習","author_link":["522855","522857","522854","522859","522856","522858"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"差分プライバシと秘密計算の融合による秘匿性がデータ提供者の数に依存しない秘匿協調学習"},{"subitem_title":"Privacy-Preserving Collaborative Learning Based on Integration of Secure Computation and Differential Privacy","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"秘匿協調学習，協調学習，秘密計算，差分プライバシ","subitem_subject_scheme":"Other"}]},"item_type_id":"18","publish_date":"2020-10-19","item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_18_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"大阪大学"},{"subitem_text_value":"大阪大学"},{"subitem_text_value":"大阪大学"}]},"item_18_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Osaka University","subitem_text_language":"en"},{"subitem_text_value":"Osaka University","subitem_text_language":"en"},{"subitem_text_value":"Osaka University","subitem_text_language":"en"}]},"item_publisher":{"attribute_name":"出版者","attribute_value_mlt":[{"subitem_publisher":"情報処理学会","subitem_publisher_language":"ja"}]},"publish_status":"0","weko_shared_id":-1,"item_file_price":{"attribute_name":"Billing file","attribute_type":"file","attribute_value_mlt":[{"url":{"url":"https://ipsj.ixsq.nii.ac.jp/record/208454/files/IPSJCSS2020026.pdf","label":"IPSJCSS2020026.pdf"},"date":[{"dateType":"Available","dateValue":"2022-10-19"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJCSS2020026.pdf","filesize":[{"value":"728.1 kB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"30"},{"tax":["include_tax"],"price":"0","billingrole":"46"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"319ef8a9-a9e2-4b6d-a797-80c82f08e139","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2020 by the Information Processing Society of Japan"}]},"item_18_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"岩花, 一輝"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"矢内, 直人"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"藤原, 融"}],"nameIdentifiers":[{}]}]},"item_18_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Kazuki, Iwahana","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Naoto, Yanai","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Toru, Fujiwara","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_5794","resourcetype":"conference paper"}]},"item_18_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"人数でデータを出し合いながら学習を行う協調学習において,差分プライバシと秘密計算の2つを融合した秘匿協調学習が近年注目されている.しかし,既存の方式ではデータ提供者の数に比例して秘匿性が保証されなくなる問題がある.本稿ではデータ提供者の数に依存せず,秘匿性と精度を両立できる融合方式SPGCを提案する. SPGCでは秘密分散で保護されたデータに対し,秘匿回路の中で差分プライバシのノイズを生成しながら勾配を計算することで,秘匿性と精度を両立する.また,実証実験では,学術用データセットにMNIST,医療用データセットにCancerおよびDiabetesを用いてSPGCの学習性能をそれぞれ評価した.とくにCancerでは,学習時間は約10時間,精度は92.2%であった.これは局所差分プライバシに基づく手法と比較して精度が約5.6%上回っている.","subitem_description_type":"Other"}]},"item_18_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"For collaborative learning which allows plural data owners to share data for the training, privacy-preserving learning protocols based on secure computation and differential privacy have been proposed in recent years. However, the confidentiality of training data in the existing protocols are downgraded in proportion to the number of data owners. In this paper, we propose a new integration protocol whose confidentiality is independent of the number of data owners. Loosely speaking, our protocol is able to guarantee both the confidentiality and accuracy by distributing data via secret sharing and generating noise within garbled circuits. <br>We also conduct experiments to evaluate the accuracy and the training time with the MNIST dataset as an academic benchmark, and the Cancer and Diabetes datasets as medical diagnosis benchmarks.","subitem_description_type":"Other"}]},"item_18_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"190","bibliographic_titles":[{"bibliographic_title":"コンピュータセキュリティシンポジウム2020論文集"}],"bibliographicPageStart":"183","bibliographicIssueDates":{"bibliographicIssueDate":"2020-10-19","bibliographicIssueDateType":"Issued"}}]},"relation_version_is_last":true,"weko_creator_id":"44499"}}