Hardening against adversarial examples with the smooth gradient method

Commonly used methods in deep learning do not utilise transformations of the residual gradient available at the inputs to update the representation in the dataset. It has been shown that this residual gradient, which can be interpreted as the first-order gradient of the input sensitivity at a partic...

Full description

Saved in:
Bibliographic Details
Published in:Soft computing (Berlin, Germany) Germany), 2018-05, Vol.22 (10), p.3203-3213
Main Authors: Mosca, Alan, Magoulas, George D.
Format: Article
Language:English
Subjects:
Artificial Intelligence
Back propagation
Computational Intelligence
Control
Deep learning
Engineering
Focus
Mathematical Logic and Foundations
Mechatronics
Neural networks
Propagation
Robotics
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by cdi_FETCH-LOGICAL-c316t-a99ee0d0a708e2486918bfc51902341bdea67b5d03a2e2d9f41358697401fdb73
cites cdi_FETCH-LOGICAL-c316t-a99ee0d0a708e2486918bfc51902341bdea67b5d03a2e2d9f41358697401fdb73
container_end_page 3213
container_issue 10
container_start_page 3203
container_title Soft computing (Berlin, Germany)
container_volume 22
creator Mosca, Alan
Magoulas, George D.
description Commonly used methods in deep learning do not utilise transformations of the residual gradient available at the inputs to update the representation in the dataset. It has been shown that this residual gradient, which can be interpreted as the first-order gradient of the input sensitivity at a particular point, may be used to improve generalisation in feed-forward neural networks, including fully connected and convolutional layers. We explore how these input gradients are related to input perturbations used to generate adversarial examples and how the networks that are trained with this technique are more robust to attacks generated with the fast gradient sign method.
doi_str_mv 10.1007/s00500-017-2998-4
format article
fullrecord <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_journals_2917933234</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2917933234</sourcerecordid><originalsourceid>FETCH-LOGICAL-c316t-a99ee0d0a708e2486918bfc51902341bdea67b5d03a2e2d9f41358697401fdb73</originalsourceid><addsrcrecordid>eNp1kMtOwzAQRS0EEqXwAewssTaMH4njJaooRarEBtaWE0_SVE1SbJfH35MSJFas5i7OvSMdQq453HIAfRcBMgAGXDNhTMHUCZlxJSXTSpvTnyyYzpU8JxcxbgEE15mckeXKBY992zfUNa7tY6LOv2OILrRuR_HTdfsdRvrRpg1NG6SxG4YxNsH5FvtEO0ybwV-Ss9rtIl793jl5XT68LFZs_fz4tLhfs0ryPDFnDCJ4cBoKFKrIDS_Kusq4ASEVLz26XJeZB-kECm9qxWU2UloBr32p5ZzcTLv7MLwdMCa7HQ6hH19aYbg2Uo47I8UnqgpDjAFruw9t58KX5WCPuuyky4667FGXPXbE1Ikj2zcY_pb_L30DDHtswA</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2917933234</pqid></control><display><type>article</type><title>Hardening against adversarial examples with the smooth gradient method</title><source>Springer Link</source><creator>Mosca, Alan ; Magoulas, George D.</creator><creatorcontrib>Mosca, Alan ; Magoulas, George D.</creatorcontrib><description>Commonly used methods in deep learning do not utilise transformations of the residual gradient available at the inputs to update the representation in the dataset. It has been shown that this residual gradient, which can be interpreted as the first-order gradient of the input sensitivity at a particular point, may be used to improve generalisation in feed-forward neural networks, including fully connected and convolutional layers. We explore how these input gradients are related to input perturbations used to generate adversarial examples and how the networks that are trained with this technique are more robust to attacks generated with the fast gradient sign method.</description><identifier>ISSN: 1432-7643</identifier><identifier>EISSN: 1433-7479</identifier><identifier>DOI: 10.1007/s00500-017-2998-4</identifier><language>eng</language><publisher>Berlin/Heidelberg: Springer Berlin Heidelberg</publisher><subject>Artificial Intelligence ; Back propagation ; Computational Intelligence ; Control ; Deep learning ; Engineering ; Focus ; Mathematical Logic and Foundations ; Mechatronics ; Neural networks ; Propagation ; Robotics</subject><ispartof>Soft computing (Berlin, Germany), 2018-05, Vol.22 (10), p.3203-3213</ispartof><rights>Springer-Verlag GmbH Germany, part of Springer Nature 2018</rights><rights>Springer-Verlag GmbH Germany, part of Springer Nature 2018.</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c316t-a99ee0d0a708e2486918bfc51902341bdea67b5d03a2e2d9f41358697401fdb73</citedby><cites>FETCH-LOGICAL-c316t-a99ee0d0a708e2486918bfc51902341bdea67b5d03a2e2d9f41358697401fdb73</cites><orcidid>0000-0001-8918-2389</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>314,776,780,27901,27902</link.rule.ids></links><search><creatorcontrib>Mosca, Alan</creatorcontrib><creatorcontrib>Magoulas, George D.</creatorcontrib><title>Hardening against adversarial examples with the smooth gradient method</title><title>Soft computing (Berlin, Germany)</title><addtitle>Soft Comput</addtitle><description>Commonly used methods in deep learning do not utilise transformations of the residual gradient available at the inputs to update the representation in the dataset. It has been shown that this residual gradient, which can be interpreted as the first-order gradient of the input sensitivity at a particular point, may be used to improve generalisation in feed-forward neural networks, including fully connected and convolutional layers. We explore how these input gradients are related to input perturbations used to generate adversarial examples and how the networks that are trained with this technique are more robust to attacks generated with the fast gradient sign method.</description><subject>Artificial Intelligence</subject><subject>Back propagation</subject><subject>Computational Intelligence</subject><subject>Control</subject><subject>Deep learning</subject><subject>Engineering</subject><subject>Focus</subject><subject>Mathematical Logic and Foundations</subject><subject>Mechatronics</subject><subject>Neural networks</subject><subject>Propagation</subject><subject>Robotics</subject><issn>1432-7643</issn><issn>1433-7479</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2018</creationdate><recordtype>article</recordtype><recordid>eNp1kMtOwzAQRS0EEqXwAewssTaMH4njJaooRarEBtaWE0_SVE1SbJfH35MSJFas5i7OvSMdQq453HIAfRcBMgAGXDNhTMHUCZlxJSXTSpvTnyyYzpU8JxcxbgEE15mckeXKBY992zfUNa7tY6LOv2OILrRuR_HTdfsdRvrRpg1NG6SxG4YxNsH5FvtEO0ybwV-Ss9rtIl793jl5XT68LFZs_fz4tLhfs0ryPDFnDCJ4cBoKFKrIDS_Kusq4ASEVLz26XJeZB-kECm9qxWU2UloBr32p5ZzcTLv7MLwdMCa7HQ6hH19aYbg2Uo47I8UnqgpDjAFruw9t58KX5WCPuuyky4667FGXPXbE1Ikj2zcY_pb_L30DDHtswA</recordid><startdate>20180501</startdate><enddate>20180501</enddate><creator>Mosca, Alan</creator><creator>Magoulas, George D.</creator><general>Springer Berlin Heidelberg</general><general>Springer Nature B.V</general><scope>AAYXX</scope><scope>CITATION</scope><scope>8FE</scope><scope>8FG</scope><scope>AFKRA</scope><scope>ARAPS</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>GNUQQ</scope><scope>HCIFZ</scope><scope>JQ2</scope><scope>K7-</scope><scope>P5Z</scope><scope>P62</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><orcidid>https://orcid.org/0000-0001-8918-2389</orcidid></search><sort><creationdate>20180501</creationdate><title>Hardening against adversarial examples with the smooth gradient method</title><author>Mosca, Alan ; Magoulas, George D.</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c316t-a99ee0d0a708e2486918bfc51902341bdea67b5d03a2e2d9f41358697401fdb73</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2018</creationdate><topic>Artificial Intelligence</topic><topic>Back propagation</topic><topic>Computational Intelligence</topic><topic>Control</topic><topic>Deep learning</topic><topic>Engineering</topic><topic>Focus</topic><topic>Mathematical Logic and Foundations</topic><topic>Mechatronics</topic><topic>Neural networks</topic><topic>Propagation</topic><topic>Robotics</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Mosca, Alan</creatorcontrib><creatorcontrib>Magoulas, George D.</creatorcontrib><collection>CrossRef</collection><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>ProQuest Central</collection><collection>Advanced Technologies &amp; Aerospace Collection</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central</collection><collection>ProQuest Central Student</collection><collection>SciTech Premium Collection (Proquest) (PQ_SDU_P3)</collection><collection>ProQuest Computer Science Collection</collection><collection>Computer Science Database</collection><collection>Advanced Technologies &amp; Aerospace Database</collection><collection>ProQuest Advanced Technologies &amp; Aerospace Collection</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><jtitle>Soft computing (Berlin, Germany)</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Mosca, Alan</au><au>Magoulas, George D.</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Hardening against adversarial examples with the smooth gradient method</atitle><jtitle>Soft computing (Berlin, Germany)</jtitle><stitle>Soft Comput</stitle><date>2018-05-01</date><risdate>2018</risdate><volume>22</volume><issue>10</issue><spage>3203</spage><epage>3213</epage><pages>3203-3213</pages><issn>1432-7643</issn><eissn>1433-7479</eissn><abstract>Commonly used methods in deep learning do not utilise transformations of the residual gradient available at the inputs to update the representation in the dataset. It has been shown that this residual gradient, which can be interpreted as the first-order gradient of the input sensitivity at a particular point, may be used to improve generalisation in feed-forward neural networks, including fully connected and convolutional layers. We explore how these input gradients are related to input perturbations used to generate adversarial examples and how the networks that are trained with this technique are more robust to attacks generated with the fast gradient sign method.</abstract><cop>Berlin/Heidelberg</cop><pub>Springer Berlin Heidelberg</pub><doi>10.1007/s00500-017-2998-4</doi><tpages>11</tpages><orcidid>https://orcid.org/0000-0001-8918-2389</orcidid></addata></record>
fulltext fulltext
identifier ISSN: 1432-7643
ispartof Soft computing (Berlin, Germany), 2018-05, Vol.22 (10), p.3203-3213
issn 1432-7643
1433-7479
language eng
recordid cdi_proquest_journals_2917933234
source Springer Link
subjects Artificial Intelligence
Back propagation
Computational Intelligence
Control
Deep learning
Engineering
Focus
Mathematical Logic and Foundations
Mechatronics
Neural networks
Propagation
Robotics
title Hardening against adversarial examples with the smooth gradient method
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-06T15%3A53%3A44IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Hardening%20against%20adversarial%20examples%20with%20the%20smooth%20gradient%20method&rft.jtitle=Soft%20computing%20(Berlin,%20Germany)&rft.au=Mosca,%20Alan&rft.date=2018-05-01&rft.volume=22&rft.issue=10&rft.spage=3203&rft.epage=3213&rft.pages=3203-3213&rft.issn=1432-7643&rft.eissn=1433-7479&rft_id=info:doi/10.1007/s00500-017-2998-4&rft_dat=%3Cproquest_cross%3E2917933234%3C/proquest_cross%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c316t-a99ee0d0a708e2486918bfc51902341bdea67b5d03a2e2d9f41358697401fdb73%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=2917933234&rft_id=info:pmid/&rfr_iscdi=true