Recycled ADMM: Improving the Privacy and Accuracy of Distributed Algorithms

Alternating direction method of multiplier (ADMM) is a powerful method to solve decentralized convex optimization problems. In distributed settings, each node performs computation with its local data and the local results are exchanged among neighboring nodes in an iterative fashion. During this ite...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on information forensics and security 2020, Vol.15, p.1723-1734
Hauptverfasser: Zhang, Xueru, Khalili, Mohammad Mahdi, Liu, Mingyan
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 1734
container_issue
container_start_page 1723
container_title IEEE transactions on information forensics and security
container_volume 15
creator Zhang, Xueru
Khalili, Mohammad Mahdi
Liu, Mingyan
description Alternating direction method of multiplier (ADMM) is a powerful method to solve decentralized convex optimization problems. In distributed settings, each node performs computation with its local data and the local results are exchanged among neighboring nodes in an iterative fashion. During this iterative process the leakage of data privacy arises and can accumulate significantly over many iterations, making it difficult to balance the privacy-accuracy tradeoff. We propose Recycled ADMM (R-ADMM), where a linear approximation is applied to every even iteration, its solution directly calculated using only results from the previous, odd iteration. It turns out that under such a scheme, half of the updates incur no privacy loss and require much less computation compared to the conventional ADMM. Moreover, R-ADMM can be further modified (MR-ADMM) such that each node independently determines its own penalty parameter over iterations. We obtain a sufficient condition for the convergence of both algorithms and provide the privacy analysis based on objective perturbation. It can be shown that the privacy-accuracy tradeoff can be improved significantly compared with conventional ADMM.
doi_str_mv 10.1109/TIFS.2019.2947867
format Article
fullrecord <record><control><sourceid>proquest_RIE</sourceid><recordid>TN_cdi_ieee_primary_8871117</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>8871117</ieee_id><sourcerecordid>2348111408</sourcerecordid><originalsourceid>FETCH-LOGICAL-c336t-f39faf94b5638ceb0048f906636b0014e7415fa832b5b461a390c757ad4497ff3</originalsourceid><addsrcrecordid>eNo9UE1PAjEUbIwmIvoDjJdNPC_2bbv98Eb4UCJEo3huuqWFJcBiu0vCv7cbCKc3L29m3mQQegTcA8DyZT4Z__QyDLKXScoF41eoA3nOUoYzuL5gILfoLoQ1xpQCEx308W3N0WzsIukPZ7PXZLLd--pQ7pZJvbLJly8P2hwTvYt3YxrfLpVLhmWofVk0davbLCtf1qttuEc3Tm-CfTjPLvodj-aD93T6-TYZ9KepIYTVqSPSaSdpkTMijC1iFuEkZoywiIFaTiF3WpCsyAvKQBOJDc-5XlAquXOki55PvjHqX2NDrdZV43fxpcoIFQBAsYgsOLGMr0Lw1qm9L7faHxVg1Xam2s5U25k6dxY1TydNaa298IXg0ZOTf6RMZm4</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2348111408</pqid></control><display><type>article</type><title>Recycled ADMM: Improving the Privacy and Accuracy of Distributed Algorithms</title><source>IEEE Electronic Library Online</source><creator>Zhang, Xueru ; Khalili, Mohammad Mahdi ; Liu, Mingyan</creator><creatorcontrib>Zhang, Xueru ; Khalili, Mohammad Mahdi ; Liu, Mingyan</creatorcontrib><description>Alternating direction method of multiplier (ADMM) is a powerful method to solve decentralized convex optimization problems. In distributed settings, each node performs computation with its local data and the local results are exchanged among neighboring nodes in an iterative fashion. During this iterative process the leakage of data privacy arises and can accumulate significantly over many iterations, making it difficult to balance the privacy-accuracy tradeoff. We propose Recycled ADMM (R-ADMM), where a linear approximation is applied to every even iteration, its solution directly calculated using only results from the previous, odd iteration. It turns out that under such a scheme, half of the updates incur no privacy loss and require much less computation compared to the conventional ADMM. Moreover, R-ADMM can be further modified (MR-ADMM) such that each node independently determines its own penalty parameter over iterations. We obtain a sufficient condition for the convergence of both algorithms and provide the privacy analysis based on objective perturbation. It can be shown that the privacy-accuracy tradeoff can be improved significantly compared with conventional ADMM.</description><identifier>ISSN: 1556-6013</identifier><identifier>EISSN: 1556-6021</identifier><identifier>DOI: 10.1109/TIFS.2019.2947867</identifier><identifier>CODEN: ITIFA6</identifier><language>eng</language><publisher>New York: IEEE</publisher><subject>Accuracy ; ADMM ; Algorithms ; Computational geometry ; Convergence ; Convexity ; Differential privacy ; Distributed databases ; distributed learning ; Iterative methods ; Mathematical analysis ; Optimization ; Perturbation methods ; Privacy ; Tradeoffs</subject><ispartof>IEEE transactions on information forensics and security, 2020, Vol.15, p.1723-1734</ispartof><rights>Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2020</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c336t-f39faf94b5638ceb0048f906636b0014e7415fa832b5b461a390c757ad4497ff3</citedby><cites>FETCH-LOGICAL-c336t-f39faf94b5638ceb0048f906636b0014e7415fa832b5b461a390c757ad4497ff3</cites><orcidid>0000-0002-0761-5943 ; 0000-0003-3295-9200 ; 0000-0002-4223-3254</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/8871117$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,776,780,792,4010,27900,27901,27902,54733</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/8871117$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Zhang, Xueru</creatorcontrib><creatorcontrib>Khalili, Mohammad Mahdi</creatorcontrib><creatorcontrib>Liu, Mingyan</creatorcontrib><title>Recycled ADMM: Improving the Privacy and Accuracy of Distributed Algorithms</title><title>IEEE transactions on information forensics and security</title><addtitle>TIFS</addtitle><description>Alternating direction method of multiplier (ADMM) is a powerful method to solve decentralized convex optimization problems. In distributed settings, each node performs computation with its local data and the local results are exchanged among neighboring nodes in an iterative fashion. During this iterative process the leakage of data privacy arises and can accumulate significantly over many iterations, making it difficult to balance the privacy-accuracy tradeoff. We propose Recycled ADMM (R-ADMM), where a linear approximation is applied to every even iteration, its solution directly calculated using only results from the previous, odd iteration. It turns out that under such a scheme, half of the updates incur no privacy loss and require much less computation compared to the conventional ADMM. Moreover, R-ADMM can be further modified (MR-ADMM) such that each node independently determines its own penalty parameter over iterations. We obtain a sufficient condition for the convergence of both algorithms and provide the privacy analysis based on objective perturbation. It can be shown that the privacy-accuracy tradeoff can be improved significantly compared with conventional ADMM.</description><subject>Accuracy</subject><subject>ADMM</subject><subject>Algorithms</subject><subject>Computational geometry</subject><subject>Convergence</subject><subject>Convexity</subject><subject>Differential privacy</subject><subject>Distributed databases</subject><subject>distributed learning</subject><subject>Iterative methods</subject><subject>Mathematical analysis</subject><subject>Optimization</subject><subject>Perturbation methods</subject><subject>Privacy</subject><subject>Tradeoffs</subject><issn>1556-6013</issn><issn>1556-6021</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNo9UE1PAjEUbIwmIvoDjJdNPC_2bbv98Eb4UCJEo3huuqWFJcBiu0vCv7cbCKc3L29m3mQQegTcA8DyZT4Z__QyDLKXScoF41eoA3nOUoYzuL5gILfoLoQ1xpQCEx308W3N0WzsIukPZ7PXZLLd--pQ7pZJvbLJly8P2hwTvYt3YxrfLpVLhmWofVk0davbLCtf1qttuEc3Tm-CfTjPLvodj-aD93T6-TYZ9KepIYTVqSPSaSdpkTMijC1iFuEkZoywiIFaTiF3WpCsyAvKQBOJDc-5XlAquXOki55PvjHqX2NDrdZV43fxpcoIFQBAsYgsOLGMr0Lw1qm9L7faHxVg1Xam2s5U25k6dxY1TydNaa298IXg0ZOTf6RMZm4</recordid><startdate>2020</startdate><enddate>2020</enddate><creator>Zhang, Xueru</creator><creator>Khalili, Mohammad Mahdi</creator><creator>Liu, Mingyan</creator><general>IEEE</general><general>The Institute of Electrical and Electronics Engineers, Inc. (IEEE)</general><scope>97E</scope><scope>RIA</scope><scope>RIE</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7SC</scope><scope>7SP</scope><scope>7TB</scope><scope>8FD</scope><scope>FR3</scope><scope>JQ2</scope><scope>KR7</scope><scope>L7M</scope><scope>L~C</scope><scope>L~D</scope><orcidid>https://orcid.org/0000-0002-0761-5943</orcidid><orcidid>https://orcid.org/0000-0003-3295-9200</orcidid><orcidid>https://orcid.org/0000-0002-4223-3254</orcidid></search><sort><creationdate>2020</creationdate><title>Recycled ADMM: Improving the Privacy and Accuracy of Distributed Algorithms</title><author>Zhang, Xueru ; Khalili, Mohammad Mahdi ; Liu, Mingyan</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c336t-f39faf94b5638ceb0048f906636b0014e7415fa832b5b461a390c757ad4497ff3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Accuracy</topic><topic>ADMM</topic><topic>Algorithms</topic><topic>Computational geometry</topic><topic>Convergence</topic><topic>Convexity</topic><topic>Differential privacy</topic><topic>Distributed databases</topic><topic>distributed learning</topic><topic>Iterative methods</topic><topic>Mathematical analysis</topic><topic>Optimization</topic><topic>Perturbation methods</topic><topic>Privacy</topic><topic>Tradeoffs</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Zhang, Xueru</creatorcontrib><creatorcontrib>Khalili, Mohammad Mahdi</creatorcontrib><creatorcontrib>Liu, Mingyan</creatorcontrib><collection>IEEE All-Society Periodicals Package (ASPP) 2005-present</collection><collection>IEEE All-Society Periodicals Package (ASPP) 1998–Present</collection><collection>IEEE Electronic Library Online</collection><collection>CrossRef</collection><collection>Computer and Information Systems Abstracts</collection><collection>Electronics &amp; Communications Abstracts</collection><collection>Mechanical &amp; Transportation Engineering Abstracts</collection><collection>Technology Research Database</collection><collection>Engineering Research Database</collection><collection>ProQuest Computer Science Collection</collection><collection>Civil Engineering Abstracts</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>Computer and Information Systems Abstracts – Academic</collection><collection>Computer and Information Systems Abstracts Professional</collection><jtitle>IEEE transactions on information forensics and security</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Zhang, Xueru</au><au>Khalili, Mohammad Mahdi</au><au>Liu, Mingyan</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Recycled ADMM: Improving the Privacy and Accuracy of Distributed Algorithms</atitle><jtitle>IEEE transactions on information forensics and security</jtitle><stitle>TIFS</stitle><date>2020</date><risdate>2020</risdate><volume>15</volume><spage>1723</spage><epage>1734</epage><pages>1723-1734</pages><issn>1556-6013</issn><eissn>1556-6021</eissn><coden>ITIFA6</coden><abstract>Alternating direction method of multiplier (ADMM) is a powerful method to solve decentralized convex optimization problems. In distributed settings, each node performs computation with its local data and the local results are exchanged among neighboring nodes in an iterative fashion. During this iterative process the leakage of data privacy arises and can accumulate significantly over many iterations, making it difficult to balance the privacy-accuracy tradeoff. We propose Recycled ADMM (R-ADMM), where a linear approximation is applied to every even iteration, its solution directly calculated using only results from the previous, odd iteration. It turns out that under such a scheme, half of the updates incur no privacy loss and require much less computation compared to the conventional ADMM. Moreover, R-ADMM can be further modified (MR-ADMM) such that each node independently determines its own penalty parameter over iterations. We obtain a sufficient condition for the convergence of both algorithms and provide the privacy analysis based on objective perturbation. It can be shown that the privacy-accuracy tradeoff can be improved significantly compared with conventional ADMM.</abstract><cop>New York</cop><pub>IEEE</pub><doi>10.1109/TIFS.2019.2947867</doi><tpages>12</tpages><orcidid>https://orcid.org/0000-0002-0761-5943</orcidid><orcidid>https://orcid.org/0000-0003-3295-9200</orcidid><orcidid>https://orcid.org/0000-0002-4223-3254</orcidid><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier ISSN: 1556-6013
ispartof IEEE transactions on information forensics and security, 2020, Vol.15, p.1723-1734
issn 1556-6013
1556-6021
language eng
recordid cdi_ieee_primary_8871117
source IEEE Electronic Library Online
subjects Accuracy
ADMM
Algorithms
Computational geometry
Convergence
Convexity
Differential privacy
Distributed databases
distributed learning
Iterative methods
Mathematical analysis
Optimization
Perturbation methods
Privacy
Tradeoffs
title Recycled ADMM: Improving the Privacy and Accuracy of Distributed Algorithms
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-11T15%3A41%3A45IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_RIE&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Recycled%20ADMM:%20Improving%20the%20Privacy%20and%20Accuracy%20of%20Distributed%20Algorithms&rft.jtitle=IEEE%20transactions%20on%20information%20forensics%20and%20security&rft.au=Zhang,%20Xueru&rft.date=2020&rft.volume=15&rft.spage=1723&rft.epage=1734&rft.pages=1723-1734&rft.issn=1556-6013&rft.eissn=1556-6021&rft.coden=ITIFA6&rft_id=info:doi/10.1109/TIFS.2019.2947867&rft_dat=%3Cproquest_RIE%3E2348111408%3C/proquest_RIE%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2348111408&rft_id=info:pmid/&rft_ieee_id=8871117&rfr_iscdi=true