{"created":"2024-05-10T02:46:12.520452+00:00","id":2000612,"links":{},"metadata":{"_buckets":{"deposit":"838a7e73-8caf-4ab7-b60a-6d10022f99e2"},"_deposit":{"created_by":14,"id":"2000612","owner":"14","owners":[14],"pid":{"revision_id":0,"type":"depid","value":"2000612"},"status":"published"},"_oai":{"id":"oai:kyutech.repo.nii.ac.jp:02000612","sets":["8:24"]},"author_link":["24877"],"control_number":"2000612","item_21_biblio_info_6":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicIssueDates":{"bibliographicIssueDate":"2024-04-23","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"9","bibliographicPageStart":"3558","bibliographicVolumeNumber":"14","bibliographic_titles":[{"bibliographic_title":"Applied Sciences","bibliographic_titleLang":"en"}]}]},"item_21_description_4":{"attribute_name":"抄録","attribute_value_mlt":[{"subitem_description":"Large Language Models (LLMs), such as ChatGPT, encounter ‘jailbreak’ challenges, wherein safeguards are circumvented to generate ethically harmful prompts. This study introduces a straightforward black-box method for efficiently crafting jailbreak prompts that bypass LLM defenses. Our technique iteratively transforms harmful prompts into benign expressions directly utilizing the target LLM, predicated on the hypothesis that LLMs can autonomously generate expressions that evade safeguards. Through experiments conducted with ChatGPT (GPT-3.5 and GPT-4) and Gemini-Pro, our method consistently achieved an attack success rate exceeding 80% within an average of five iterations for forbidden questions and proved to be robust against model updates. The jailbreak prompts generated were not only naturally worded and succinct, but also challenging to defend against. These findings suggest that the creation of effective jailbreak prompts is less complex than previously believed, underscoring the heightened risk posed by black-box jailbreak attacks.","subitem_description_language":"en","subitem_description_type":"Abstract"}]},"item_21_link_62":{"attribute_name":"研究者情報","attribute_value_mlt":[{"subitem_link_url":"https://hyokadb02.jimu.kyutech.ac.jp/html/100000509_ja.html"}]},"item_21_publisher_7":{"attribute_name":"出版社","attribute_value_mlt":[{"subitem_publisher":"MDPI"}]},"item_21_relation_12":{"attribute_name":"DOI","attribute_value_mlt":[{"subitem_relation_type":"isIdenticalTo","subitem_relation_type_id":{"subitem_relation_type_id_text":"https://doi.org/10.3390/app14093558","subitem_relation_type_select":"DOI"}}]},"item_21_rights_13":{"attribute_name":"著作権関連情報","attribute_value_mlt":[{"subitem_rights":"Copyright (c) 2024 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).","subitem_rights_resource":"https://creativecommons.org/licenses/by/4.0/"}]},"item_21_select_59":{"attribute_name":"査読の有無","attribute_value_mlt":[{"subitem_select_item":"yes"}]},"item_21_source_id_8":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"2076-3417","subitem_source_identifier_type":"EISSN"}]},"item_21_text_28":{"attribute_name":"論文ID（連携）","attribute_value_mlt":[{"subitem_text_value":"10430409"}]},"item_21_text_63":{"attribute_name":"連携ID","attribute_value_mlt":[{"subitem_text_value":"12223"}]},"item_21_version_type_58":{"attribute_name":"出版タイプ","attribute_value_mlt":[{"subitem_version_resource":"http://purl.org/coar/version/c_970fb48d4fbd8a85","subitem_version_type":"VoR"}]},"item_creator":{"attribute_name":"著者","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Takemoto, Kazuhiro","creatorNameLang":"en"},{"creatorName":"竹本, 和広","creatorNameLang":"ja"}],"familyNames":[{},{}],"givenNames":[{},{}],"nameIdentifiers":[{"nameIdentifier":"24877","nameIdentifierScheme":"WEKO"},{"nameIdentifier":"40512356","nameIdentifierScheme":"e-Rad","nameIdentifierURI":"https://nrid.nii.ac.jp/ja/nrid/1000040512356"},{"nameIdentifier":"35270356700","nameIdentifierScheme":"Scopus著者ID","nameIdentifierURI":"https://www.scopus.com/authid/detail.uri?authorId=35270356700"},{"nameIdentifier":"0000-0002-6355-1366","nameIdentifierScheme":"ORCiD","nameIdentifierURI":"https://orcid.org/0000-0002-6355-1366"},{"nameIdentifier":"100000509","nameIdentifierScheme":"九工大研究者情報","nameIdentifierURI":"https://hyokadb02.jimu.kyutech.ac.jp/html/##_ja.html"}]}]},"item_files":{"attribute_name":"ファイル情報","attribute_type":"file","attribute_value_mlt":[{"accessrole":"open_access","date":[{"dateType":"Available","dateValue":"2024-05-10"}],"filename":"10430409.pdf","filesize":[{"value":"299 KB"}],"format":"application/pdf","mimetype":"application/pdf","url":{"url":"https://kyutech.repo.nii.ac.jp/record/2000612/files/10430409.pdf"},"version_id":"ee06509b-4915-4b23-9bac-44130d7cf65e"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"large language models","subitem_subject_scheme":"Other"},{"subitem_subject":"jailbreak attacks","subitem_subject_scheme":"Other"},{"subitem_subject":"security and privacy","subitem_subject_scheme":"Other"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"eng"}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourcetype":"journal article","resourceuri":"http://purl.org/coar/resource_type/c_6501"}]},"item_title":"All in How You Ask for It: Simple Black-Box Method for Jailbreak Attacks","item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"All in How You Ask for It: Simple Black-Box Method for Jailbreak Attacks","subitem_title_language":"en"}]},"item_type_id":"21","owner":"14","path":["24"],"pubdate":{"attribute_name":"PubDate","attribute_value":"2024-05-10"},"publish_date":"2024-05-10","publish_status":"0","recid":"2000612","relation_version_is_last":true,"title":["All in How You Ask for It: Simple Black-Box Method for Jailbreak Attacks"],"weko_creator_id":"14","weko_shared_id":-1},"updated":"2024-05-10T02:53:40.755672+00:00"}