{"created":"2025-01-19T01:20:52.862616+00:00","updated":"2025-01-19T14:23:50.722352+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00220857","sets":["6504:11035:11043"]},"path":["11043"],"owner":"44499","recid":"220857","title":["Deepfakeを破壊する摂動の転移性調査と効率的な最適化手法の検討"],"pubdate":{"attribute_name":"公開日","attribute_value":"2022-02-17"},"_buckets":{"deposit":"3a88b127-604a-44f4-982d-ba5c7543984b"},"_deposit":{"id":"220857","pid":{"type":"depid","value":"220857","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"Deepfakeを破壊する摂動の転移性調査と効率的な最適化手法の検討","author_link":["577932","577930","577931"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"Deepfakeを破壊する摂動の転移性調査と効率的な最適化手法の検討"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"人工知能と認知科学","subitem_subject_scheme":"Other"}]},"item_type_id":"22","publish_date":"2022-02-17","item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_22_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"早大"},{"subitem_text_value":"早大"},{"subitem_text_value":"早大"}]},"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/220857/files/IPSJ-Z84-4Q-02.pdf","label":"IPSJ-Z84-4Q-02.pdf"},"date":[{"dateType":"Available","dateValue":"2022-10-22"}],"format":"application/pdf","filename":"IPSJ-Z84-4Q-02.pdf","filesize":[{"value":"1.1 MB"}],"mimetype":"application/pdf","accessrole":"open_date","version_id":"91bb9caf-4630-4d58-aa67-79f501db2421","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2022 by the Information Processing Society of Japan"}]},"item_22_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"加藤, 義道"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"福原, 吉博"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"森島, 繁生"}],"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_22_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN00349328","subitem_source_identifier_type":"NCID"}]},"item_22_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"Deepfakeは深層学習を用いたメディア合成技術である。これにより、人の印象を悪くするような悪意ある偽動画が作成され問題になっている。これを解決するために、人が認識できない微弱な摂動を用いてDNN変換モデルを破壊する手法が注目されている。既存手法では、最適化したモデルに対しては効率的な破壊がされていたが、異なるモデルへの転移性の調査は行われていなかった。我々は、複数の変換モデルにおける摂動の転移性を網羅的に調査した。既存手法ではノイズを大きくすることである程度の転移を確認したが、大きな摂動を加えることで画像の品質が低下した。これを踏まえて、大きな摂動を加えても画像の品質が劣化しないような手法を検討した。","subitem_description_type":"Other"}]},"item_22_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"220","bibliographic_titles":[{"bibliographic_title":"第84回全国大会講演論文集"}],"bibliographicPageStart":"219","bibliographicIssueDates":{"bibliographicIssueDate":"2022-02-17","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"1","bibliographicVolumeNumber":"2022"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":220857,"links":{}}