Failure Modes in Machine Learning Systems

In the last two years, more than 200 papers have been written on how machine learning (ML) systems can fail because of adversarial attacks on the algorithms and data; this number balloons if we were to incorporate papers covering non-adversarial failure modes. The spate of papers has made it difficu...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kumar, Ram Shankar Siva, Brien, David O, Albert, Kendra, Viljöen, Salomé, Snover, Jeffrey
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 Kumar, Ram Shankar Siva
Brien, David O
Albert, Kendra
Viljöen, Salomé
Snover, Jeffrey
description In the last two years, more than 200 papers have been written on how machine learning (ML) systems can fail because of adversarial attacks on the algorithms and data; this number balloons if we were to incorporate papers covering non-adversarial failure modes. The spate of papers has made it difficult for ML practitioners, let alone engineers, lawyers, and policymakers, to keep up with the attacks against and defenses of ML systems. However, as these systems become more pervasive, the need to understand how they fail, whether by the hand of an adversary or due to the inherent design of a system, will only become more pressing. In order to equip software developers, security incident responders, lawyers, and policy makers with a common vernacular to talk about this problem, we developed a framework to classify failures into "Intentional failures" where the failure is caused by an active adversary attempting to subvert the system to attain her goals; and "Unintentional failures" where the failure is because an ML system produces an inherently unsafe outcome. After developing the initial version of the taxonomy last year, we worked with security and ML teams across Microsoft, 23 external partners, standards organization, and governments to understand how stakeholders would use our framework. Throughout the paper, we attempt to highlight how machine learning failure modes are meaningfully different from traditional software failures from a technology and policy perspective.
doi_str_mv 10.48550/arxiv.1911.11034
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_1911_11034</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>1911_11034</sourcerecordid><originalsourceid>FETCH-LOGICAL-a674-27479b0bab5080c5c05b9cf9e725d10402aee55e2d3910a9175f31eb5f55b8423</originalsourceid><addsrcrecordid>eNotzj1PwzAQgGEvDKjwA5jw2iHhzvbheKwq-iGlYqB7dE4uxVIbKgdQ--8Rbad3e_Uo9YRQuooIXjif0m-JAbFEBOvu1XTBaf-TRW--Ohl1GvSG2880iK6F85CGnf44j99yGB_UXc_7UR5vnajt4m07XxX1-3I9n9UFv3pXGO98iBA5ElTQUgsUQ9sH8YY6BAeGRYjEdDYgcEBPvUWJ1BPFyhk7Uc_X7cXaHHM6cD43_-bmYrZ_trM6Qg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Failure Modes in Machine Learning Systems</title><source>arXiv.org</source><creator>Kumar, Ram Shankar Siva ; Brien, David O ; Albert, Kendra ; Viljöen, Salomé ; Snover, Jeffrey</creator><creatorcontrib>Kumar, Ram Shankar Siva ; Brien, David O ; Albert, Kendra ; Viljöen, Salomé ; Snover, Jeffrey</creatorcontrib><description>In the last two years, more than 200 papers have been written on how machine learning (ML) systems can fail because of adversarial attacks on the algorithms and data; this number balloons if we were to incorporate papers covering non-adversarial failure modes. The spate of papers has made it difficult for ML practitioners, let alone engineers, lawyers, and policymakers, to keep up with the attacks against and defenses of ML systems. However, as these systems become more pervasive, the need to understand how they fail, whether by the hand of an adversary or due to the inherent design of a system, will only become more pressing. In order to equip software developers, security incident responders, lawyers, and policy makers with a common vernacular to talk about this problem, we developed a framework to classify failures into "Intentional failures" where the failure is caused by an active adversary attempting to subvert the system to attain her goals; and "Unintentional failures" where the failure is because an ML system produces an inherently unsafe outcome. After developing the initial version of the taxonomy last year, we worked with security and ML teams across Microsoft, 23 external partners, standards organization, and governments to understand how stakeholders would use our framework. Throughout the paper, we attempt to highlight how machine learning failure modes are meaningfully different from traditional software failures from a technology and policy perspective.</description><identifier>DOI: 10.48550/arxiv.1911.11034</identifier><language>eng</language><subject>Computer Science - Cryptography and Security ; Computer Science - Learning ; Statistics - Machine Learning</subject><creationdate>2019-11</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,781,886</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/1911.11034$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.1911.11034$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Kumar, Ram Shankar Siva</creatorcontrib><creatorcontrib>Brien, David O</creatorcontrib><creatorcontrib>Albert, Kendra</creatorcontrib><creatorcontrib>Viljöen, Salomé</creatorcontrib><creatorcontrib>Snover, Jeffrey</creatorcontrib><title>Failure Modes in Machine Learning Systems</title><description>In the last two years, more than 200 papers have been written on how machine learning (ML) systems can fail because of adversarial attacks on the algorithms and data; this number balloons if we were to incorporate papers covering non-adversarial failure modes. The spate of papers has made it difficult for ML practitioners, let alone engineers, lawyers, and policymakers, to keep up with the attacks against and defenses of ML systems. However, as these systems become more pervasive, the need to understand how they fail, whether by the hand of an adversary or due to the inherent design of a system, will only become more pressing. In order to equip software developers, security incident responders, lawyers, and policy makers with a common vernacular to talk about this problem, we developed a framework to classify failures into "Intentional failures" where the failure is caused by an active adversary attempting to subvert the system to attain her goals; and "Unintentional failures" where the failure is because an ML system produces an inherently unsafe outcome. After developing the initial version of the taxonomy last year, we worked with security and ML teams across Microsoft, 23 external partners, standards organization, and governments to understand how stakeholders would use our framework. Throughout the paper, we attempt to highlight how machine learning failure modes are meaningfully different from traditional software failures from a technology and policy perspective.</description><subject>Computer Science - Cryptography and Security</subject><subject>Computer Science - Learning</subject><subject>Statistics - Machine Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2019</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotzj1PwzAQgGEvDKjwA5jw2iHhzvbheKwq-iGlYqB7dE4uxVIbKgdQ--8Rbad3e_Uo9YRQuooIXjif0m-JAbFEBOvu1XTBaf-TRW--Ohl1GvSG2880iK6F85CGnf44j99yGB_UXc_7UR5vnajt4m07XxX1-3I9n9UFv3pXGO98iBA5ElTQUgsUQ9sH8YY6BAeGRYjEdDYgcEBPvUWJ1BPFyhk7Uc_X7cXaHHM6cD43_-bmYrZ_trM6Qg</recordid><startdate>20191125</startdate><enddate>20191125</enddate><creator>Kumar, Ram Shankar Siva</creator><creator>Brien, David O</creator><creator>Albert, Kendra</creator><creator>Viljöen, Salomé</creator><creator>Snover, Jeffrey</creator><scope>AKY</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20191125</creationdate><title>Failure Modes in Machine Learning Systems</title><author>Kumar, Ram Shankar Siva ; Brien, David O ; Albert, Kendra ; Viljöen, Salomé ; Snover, Jeffrey</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a674-27479b0bab5080c5c05b9cf9e725d10402aee55e2d3910a9175f31eb5f55b8423</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2019</creationdate><topic>Computer Science - Cryptography and Security</topic><topic>Computer Science - Learning</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Kumar, Ram Shankar Siva</creatorcontrib><creatorcontrib>Brien, David O</creatorcontrib><creatorcontrib>Albert, Kendra</creatorcontrib><creatorcontrib>Viljöen, Salomé</creatorcontrib><creatorcontrib>Snover, Jeffrey</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Kumar, Ram Shankar Siva</au><au>Brien, David O</au><au>Albert, Kendra</au><au>Viljöen, Salomé</au><au>Snover, Jeffrey</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Failure Modes in Machine Learning Systems</atitle><date>2019-11-25</date><risdate>2019</risdate><abstract>In the last two years, more than 200 papers have been written on how machine learning (ML) systems can fail because of adversarial attacks on the algorithms and data; this number balloons if we were to incorporate papers covering non-adversarial failure modes. The spate of papers has made it difficult for ML practitioners, let alone engineers, lawyers, and policymakers, to keep up with the attacks against and defenses of ML systems. However, as these systems become more pervasive, the need to understand how they fail, whether by the hand of an adversary or due to the inherent design of a system, will only become more pressing. In order to equip software developers, security incident responders, lawyers, and policy makers with a common vernacular to talk about this problem, we developed a framework to classify failures into "Intentional failures" where the failure is caused by an active adversary attempting to subvert the system to attain her goals; and "Unintentional failures" where the failure is because an ML system produces an inherently unsafe outcome. After developing the initial version of the taxonomy last year, we worked with security and ML teams across Microsoft, 23 external partners, standards organization, and governments to understand how stakeholders would use our framework. Throughout the paper, we attempt to highlight how machine learning failure modes are meaningfully different from traditional software failures from a technology and policy perspective.</abstract><doi>10.48550/arxiv.1911.11034</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.1911.11034
ispartof
issn
language eng
recordid cdi_arxiv_primary_1911_11034
source arXiv.org
subjects Computer Science - Cryptography and Security
Computer Science - Learning
Statistics - Machine Learning
title Failure Modes in Machine Learning Systems
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-09T19%3A03%3A18IST&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=Failure%20Modes%20in%20Machine%20Learning%20Systems&rft.au=Kumar,%20Ram%20Shankar%20Siva&rft.date=2019-11-25&rft_id=info:doi/10.48550/arxiv.1911.11034&rft_dat=%3Carxiv_GOX%3E1911_11034%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