Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Information

We describe a simple and effective method (Spectral Attribute removaL; SAL) to remove private or guarded information from neural representations. Our method uses matrix decomposition to project the input representations into directions with reduced covariance with the guarded information rather than...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Shao, Shun, Ziser, Yftah, Cohen, Shay B
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Shao, Shun
Ziser, Yftah
Cohen, Shay B
description We describe a simple and effective method (Spectral Attribute removaL; SAL) to remove private or guarded information from neural representations. Our method uses matrix decomposition to project the input representations into directions with reduced covariance with the guarded information rather than maximal covariance as factorization methods normally use. We begin with linear information removal and proceed to generalize our algorithm to the case of nonlinear information removal using kernels. Our experiments demonstrate that our algorithm retains better main task performance after removing the guarded information compared to previous work. In addition, our experiments demonstrate that we need a relatively small amount of guarded attribute data to remove information about these attributes, which lowers the exposure to sensitive data and is more suitable for low-resource scenarios. Code is available at https://github.com/jasonshaoshun/SAL.
doi_str_mv 10.48550/arxiv.2203.07893
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2203_07893</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2203_07893</sourcerecordid><originalsourceid>FETCH-LOGICAL-a673-f3c9e73d14c2c0b599a97e36cb7945679a79e01597ee6eb683eab7db9f011ec13</originalsourceid><addsrcrecordid>eNotj7FOwzAURb0woMIHMOGNKcGOkzhmiwqEShFI0D16tl8kS05cOW6hf09pmY7uHY50CLnjLC-bqmKPEH_cIS8KJnImGyWuiemCt_Q54DI_JNr6bzgutPMuJYxP9GuHJkXw9BOncDgxjLR3M0KkMFv6HmZ_Wd0eokVL25Si0_uEdDOPIU6QXJhvyNUIfsHbf67I9vVlu37L-o9us277DGopslEYhVJYXprCMF0pBUqiqI2WqqxqqUAqZLw6nVijrhuBoKXVamSco-FiRe4v2nPlsItugngc_mqHc634BZkfUHI</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Information</title><source>arXiv.org</source><creator>Shao, Shun ; Ziser, Yftah ; Cohen, Shay B</creator><creatorcontrib>Shao, Shun ; Ziser, Yftah ; Cohen, Shay B</creatorcontrib><description>We describe a simple and effective method (Spectral Attribute removaL; SAL) to remove private or guarded information from neural representations. Our method uses matrix decomposition to project the input representations into directions with reduced covariance with the guarded information rather than maximal covariance as factorization methods normally use. We begin with linear information removal and proceed to generalize our algorithm to the case of nonlinear information removal using kernels. Our experiments demonstrate that our algorithm retains better main task performance after removing the guarded information compared to previous work. In addition, our experiments demonstrate that we need a relatively small amount of guarded attribute data to remove information about these attributes, which lowers the exposure to sensitive data and is more suitable for low-resource scenarios. Code is available at https://github.com/jasonshaoshun/SAL.</description><identifier>DOI: 10.48550/arxiv.2203.07893</identifier><language>eng</language><subject>Computer Science - Computation and Language ; Computer Science - Learning</subject><creationdate>2022-03</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2203.07893$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2203.07893$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Shao, Shun</creatorcontrib><creatorcontrib>Ziser, Yftah</creatorcontrib><creatorcontrib>Cohen, Shay B</creatorcontrib><title>Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Information</title><description>We describe a simple and effective method (Spectral Attribute removaL; SAL) to remove private or guarded information from neural representations. Our method uses matrix decomposition to project the input representations into directions with reduced covariance with the guarded information rather than maximal covariance as factorization methods normally use. We begin with linear information removal and proceed to generalize our algorithm to the case of nonlinear information removal using kernels. Our experiments demonstrate that our algorithm retains better main task performance after removing the guarded information compared to previous work. In addition, our experiments demonstrate that we need a relatively small amount of guarded attribute data to remove information about these attributes, which lowers the exposure to sensitive data and is more suitable for low-resource scenarios. Code is available at https://github.com/jasonshaoshun/SAL.</description><subject>Computer Science - Computation and Language</subject><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj7FOwzAURb0woMIHMOGNKcGOkzhmiwqEShFI0D16tl8kS05cOW6hf09pmY7uHY50CLnjLC-bqmKPEH_cIS8KJnImGyWuiemCt_Q54DI_JNr6bzgutPMuJYxP9GuHJkXw9BOncDgxjLR3M0KkMFv6HmZ_Wd0eokVL25Si0_uEdDOPIU6QXJhvyNUIfsHbf67I9vVlu37L-o9us277DGopslEYhVJYXprCMF0pBUqiqI2WqqxqqUAqZLw6nVijrhuBoKXVamSco-FiRe4v2nPlsItugngc_mqHc634BZkfUHI</recordid><startdate>20220315</startdate><enddate>20220315</enddate><creator>Shao, Shun</creator><creator>Ziser, Yftah</creator><creator>Cohen, Shay B</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20220315</creationdate><title>Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Information</title><author>Shao, Shun ; Ziser, Yftah ; Cohen, Shay B</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a673-f3c9e73d14c2c0b599a97e36cb7945679a79e01597ee6eb683eab7db9f011ec13</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Computer Science - Computation and Language</topic><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Shao, Shun</creatorcontrib><creatorcontrib>Ziser, Yftah</creatorcontrib><creatorcontrib>Cohen, Shay B</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Shao, Shun</au><au>Ziser, Yftah</au><au>Cohen, Shay B</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Information</atitle><date>2022-03-15</date><risdate>2022</risdate><abstract>We describe a simple and effective method (Spectral Attribute removaL; SAL) to remove private or guarded information from neural representations. Our method uses matrix decomposition to project the input representations into directions with reduced covariance with the guarded information rather than maximal covariance as factorization methods normally use. We begin with linear information removal and proceed to generalize our algorithm to the case of nonlinear information removal using kernels. Our experiments demonstrate that our algorithm retains better main task performance after removing the guarded information compared to previous work. In addition, our experiments demonstrate that we need a relatively small amount of guarded attribute data to remove information about these attributes, which lowers the exposure to sensitive data and is more suitable for low-resource scenarios. Code is available at https://github.com/jasonshaoshun/SAL.</abstract><doi>10.48550/arxiv.2203.07893</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2203.07893
ispartof
issn
language eng
recordid cdi_arxiv_primary_2203_07893
source arXiv.org
subjects Computer Science - Computation and Language
Computer Science - Learning
title Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Information
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-11T19%3A06%3A46IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Gold%20Doesn't%20Always%20Glitter:%20Spectral%20Removal%20of%20Linear%20and%20Nonlinear%20Guarded%20Attribute%20Information&rft.au=Shao,%20Shun&rft.date=2022-03-15&rft_id=info:doi/10.48550/arxiv.2203.07893&rft_dat=%3Carxiv_GOX%3E2203_07893%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true