GAN图像对抗样本生成方法
TP393; 为了提高生成对抗网络模型对抗样本的多样性和攻击成功率,提出了一种GAN图像对抗样本生成方法.首先,利用原始样本集整体训练一个深度卷积对抗生成网络G1,模拟原始样本集分布;其次,在黑盒攻击场景下,利用模型蒸馏方法对目标模型进行黑盒复制,获取目标模型的本地复制;然后以G1的输出作为输入,以蒸馏模型作为目标模型,训练生成对抗网络G2,在有目标攻击情况下还需输入目标类别,G2用以生成输入数据针对目标类别的扰动;最后将样本与扰动相加并以像素灰度值区间进行规范化,得到对抗样本.实验结果表明,在相同输入条件下该方法产生图像对抗样本平均SSIM指标、MI指标和Cosin相似度分别降低50.7％、...
Saved in:
| Published in | 计算机科学与探索 Vol. 15; no. 4; pp. 702 - 711 |
|---|---|
| Main Authors | , , |
| Format | Journal Article |
| Language | Chinese |
| Published |
西安邮电大学 计算机学院,西安 710121
01.04.2021
|
| Subjects | |
| Online Access | Get full text |
| ISSN | 1673-9418 |
| DOI | 10.3778/j.issn.1673-9418.2005022 |
Cover
| Abstract | TP393; 为了提高生成对抗网络模型对抗样本的多样性和攻击成功率,提出了一种GAN图像对抗样本生成方法.首先,利用原始样本集整体训练一个深度卷积对抗生成网络G1,模拟原始样本集分布;其次,在黑盒攻击场景下,利用模型蒸馏方法对目标模型进行黑盒复制,获取目标模型的本地复制;然后以G1的输出作为输入,以蒸馏模型作为目标模型,训练生成对抗网络G2,在有目标攻击情况下还需输入目标类别,G2用以生成输入数据针对目标类别的扰动;最后将样本与扰动相加并以像素灰度值区间进行规范化,得到对抗样本.实验结果表明,在相同输入条件下该方法产生图像对抗样本平均SSIM指标、MI指标和Cosin相似度分别降低50.7％、10.96％和28.7％,平均均方误差值(MSE)和图像指纹的海明距离分别提升7.6％和1974.80,同时MNIST数据集和CIFAR10数据集下模型平均攻击成功率在95％以上. |
|---|---|
| AbstractList | TP393; 为了提高生成对抗网络模型对抗样本的多样性和攻击成功率,提出了一种GAN图像对抗样本生成方法.首先,利用原始样本集整体训练一个深度卷积对抗生成网络G1,模拟原始样本集分布;其次,在黑盒攻击场景下,利用模型蒸馏方法对目标模型进行黑盒复制,获取目标模型的本地复制;然后以G1的输出作为输入,以蒸馏模型作为目标模型,训练生成对抗网络G2,在有目标攻击情况下还需输入目标类别,G2用以生成输入数据针对目标类别的扰动;最后将样本与扰动相加并以像素灰度值区间进行规范化,得到对抗样本.实验结果表明,在相同输入条件下该方法产生图像对抗样本平均SSIM指标、MI指标和Cosin相似度分别降低50.7％、10.96％和28.7％,平均均方误差值(MSE)和图像指纹的海明距离分别提升7.6％和1974.80,同时MNIST数据集和CIFAR10数据集下模型平均攻击成功率在95％以上. |
| Author | 王曙燕 孙家泽 金航 |
| AuthorAffiliation | 西安邮电大学 计算机学院,西安 710121 |
| AuthorAffiliation_xml | – name: 西安邮电大学 计算机学院,西安 710121 |
| Author_FL | SUN Jiaze WANG Shuyan JIN Hang |
| Author_FL_xml | – sequence: 1 fullname: WANG Shuyan – sequence: 2 fullname: JIN Hang – sequence: 3 fullname: SUN Jiaze |
| Author_xml | – sequence: 1 fullname: 王曙燕 – sequence: 2 fullname: 金航 – sequence: 3 fullname: 孙家泽 |
| BookMark | eNo9jb1KQzEYQDNUsNa-g44O9_olX-5NMpaiVSi66Fy-mybSq6RgFHV2ERcd_BkU9AEEBwfp4tvc-BoqitOBM5yzwFphGhxjSxxyVEqv1vkkxpDzUmFmJNe5AChAiBZr_7t51o1xUkEhpeCq1G22POhtNQ8fzflV8zpLl_fp-T09vnzePKWL63Q3S2-3i2zO00F03T922O762k5_IxtuDzb7vWEWOUjIvLZjX0JlhR8jqLFWjmPlrXJKWo6oDHlnSRsshSHglqQtiIwRDr1Bwg5b-e2eUPAU9kb19PgwfB9Hdaz3T8-OogDxs-KAX_YcT_0 |
| ClassificationCodes | TP393 |
| ContentType | Journal Article |
| Copyright | Copyright © Wanfang Data Co. Ltd. All Rights Reserved. |
| Copyright_xml | – notice: Copyright © Wanfang Data Co. Ltd. All Rights Reserved. |
| DBID | 2B. 4A8 92I 93N PSX TCJ |
| DOI | 10.3778/j.issn.1673-9418.2005022 |
| DatabaseName | Wanfang Data Journals - Hong Kong WANFANG Data Centre Wanfang Data Journals 万方数据期刊 - 香港版 China Online Journals (COJ) China Online Journals (COJ) |
| DatabaseTitleList | |
| DeliveryMethod | fulltext_linktorsrc |
| DocumentTitle_FL | Method for Image Adversarial Samples Generating Based on GAN |
| EndPage | 711 |
| ExternalDocumentID | jsjkxyts202104010 |
| GrantInformation_xml | – fundername: (陕西省重点研发计划项目); (西安市科技计划项目); (西安邮电大学研究生创新基金项目) funderid: (陕西省重点研发计划项目); (西安市科技计划项目); (西安邮电大学研究生创新基金项目) |
| GroupedDBID | 2B. 4A8 92I 93N ALMA_UNASSIGNED_HOLDINGS M~E PSX TCJ |
| ID | FETCH-LOGICAL-s1040-f8cdf60bc2fd307d87e13bfc7e74c13379afeca893629a01ca4c5aa992e3f93a3 |
| ISSN | 1673-9418 |
| IngestDate | Thu May 29 04:00:17 EDT 2025 |
| IsPeerReviewed | true |
| IsScholarly | true |
| Issue | 4 |
| Keywords | 神经网络 生成对抗网络(GAN) 对抗样本 模型蒸馏 图像多样性 |
| Language | Chinese |
| LinkModel | OpenURL |
| MergedId | FETCHMERGED-LOGICAL-s1040-f8cdf60bc2fd307d87e13bfc7e74c13379afeca893629a01ca4c5aa992e3f93a3 |
| PageCount | 10 |
| ParticipantIDs | wanfang_journals_jsjkxyts202104010 |
| PublicationCentury | 2000 |
| PublicationDate | 2021-04-01 |
| PublicationDateYYYYMMDD | 2021-04-01 |
| PublicationDate_xml | – month: 04 year: 2021 text: 2021-04-01 day: 01 |
| PublicationDecade | 2020 |
| PublicationTitle | 计算机科学与探索 |
| PublicationTitle_FL | Journal of Frontiers of Computer Science & Technology |
| PublicationYear | 2021 |
| Publisher | 西安邮电大学 计算机学院,西安 710121 |
| Publisher_xml | – name: 西安邮电大学 计算机学院,西安 710121 |
| SSID | ssib054421768 ssib002040941 ssib002423894 ssib051375751 ssib023646573 ssib036438069 ssib002040926 |
| Score | 2.3065524 |
| Snippet | TP393; 为了提高生成对抗网络模型对抗样本的多样性和攻击成功率,提出了一种GAN图像对抗样本生成方法.首先,利用原始样本集整体训练一个深度卷积对抗生成网络G1,模拟原始样... |
| SourceID | wanfang |
| SourceType | Aggregation Database |
| StartPage | 702 |
| Title | GAN图像对抗样本生成方法 |
| URI | https://d.wanfangdata.com.cn/periodical/jsjkxyts202104010 |
| Volume | 15 |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| journalDatabaseRights | – providerCode: PRVEBS databaseName: Inspec with Full Text issn: 1673-9418 databaseCode: ADMLS dateStart: 20200501 customDbUrl: isFulltext: true dateEnd: 99991231 titleUrlDefault: https://www.ebsco.com/products/research-databases/inspec-full-text omitProxy: false ssIdentifier: ssib002423894 providerName: EBSCOhost – providerCode: PRVHPJ databaseName: ROAD: Directory of Open Access Scholarly Resources issn: 1673-9418 databaseCode: M~E dateStart: 20070101 customDbUrl: isFulltext: true dateEnd: 99991231 titleUrlDefault: https://road.issn.org omitProxy: true ssIdentifier: ssib054421768 providerName: ISSN International Centre |
| link | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV1Lb9QwEI5KuXBBIEC8igpiTlVKEtuxfbS3WSqk9tRKvVVJNgEVaZHYrQQ9cOGCuMCBxwEk-AFIHDigXvg3G_4GM85jAwVUuETj8WM8ntjzOetZe95NLUqeISzw-Uhwn-tU-JqJwmfoXnE7gS5RUjTyxma8vs3v7IidhRNLvVNL-9NsNT_4bVzJ_1gVeWhXipL9B8t2jSIDabQvPtHC-DyWjW-bTUgEaAs2IUIxUEMizBCshiQGZUBLIkwA1hF6AGYAiQTNQQ9dGQU6cFlxU8syqO-kbGErJApMAiakikjotilrHAdbC53cNTAxJBysApW4xrFWRGUsR6I1LzEwR1nXjAWtHUe2Yl0R7RghycYuGjPPcYKojqDO2Ljps13rf8OIwt7RF3rrnBIC7LCpp7STERFdj4cVLouTQq02K3_VvVNZU39Q02hwVMyKpL83C3vrfyyZr3njEloHIXoTgfdWexlEPeAga6_xq09iUirnk0jAaifAfc4Lomjuh7vTkXuTvfuPHk8nNFK4xlIA4ckInRbdTLLxJJmjK8zU_d0hpflPYcoIR7vllq4KiMUcrWKSqSDu0KwImaRf4bo057hfrYNJ217XZ-FIpVt_UshFv43LdHy3B9S2zninmx3Wsqmny1lv4eDeOe86TpXZu2-zpy9mnw-r52-rj1-r95--v_pQPXtZvTmsvrw-720Pk63But_cDuJPaFT8UuWjMg6yPCpH6KhGShYhy8pcFpLnIWNSp2WRp4jH40inQZinPBdpqnVUsFKzlF3wFscPxsVFb1lzVpRZXoSFyHH7EKfRiE6LjTjhu1Bml7wbjUq7zeyf7B4x0uXjFLrinZq__1e9xenD_WIJUe00u-Zs-wO7CG6T |
| linkProvider | ISSN International Centre |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=GAN%E5%9B%BE%E5%83%8F%E5%AF%B9%E6%8A%97%E6%A0%B7%E6%9C%AC%E7%94%9F%E6%88%90%E6%96%B9%E6%B3%95&rft.jtitle=%E8%AE%A1%E7%AE%97%E6%9C%BA%E7%A7%91%E5%AD%A6%E4%B8%8E%E6%8E%A2%E7%B4%A2&rft.au=%E7%8E%8B%E6%9B%99%E7%87%95&rft.au=%E9%87%91%E8%88%AA&rft.au=%E5%AD%99%E5%AE%B6%E6%B3%BD&rft.date=2021-04-01&rft.pub=%E8%A5%BF%E5%AE%89%E9%82%AE%E7%94%B5%E5%A4%A7%E5%AD%A6+%E8%AE%A1%E7%AE%97%E6%9C%BA%E5%AD%A6%E9%99%A2%2C%E8%A5%BF%E5%AE%89+710121&rft.issn=1673-9418&rft.volume=15&rft.issue=4&rft.spage=702&rft.epage=711&rft_id=info:doi/10.3778%2Fj.issn.1673-9418.2005022&rft.externalDocID=jsjkxyts202104010 |
| thumbnail_s | http://utb.summon.serialssolutions.com/2.0.0/image/custom?url=http%3A%2F%2Fwww.wanfangdata.com.cn%2Fimages%2FPeriodicalImages%2Fjsjkxyts%2Fjsjkxyts.jpg |