An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data

Yovine, Sergio - Mayr, Franz - Sosa, Sebastián - Visca, Ramiro

Resumen:

This paper explores the use of Private Aggregation of Teacher Ensembles (PATE) in a setting where students have their own private data that cannot be revealed as is to the ensemble. We propose a privacy model that introduces a local differentially private mechanism to protect student data. We implemented and analyzed it in case studies from security and health domains, and the result of the experiment was twofold. First, this model does not significantly affecs predictive capabilities, and second, it unveiled interesting issues with the so-called data dependency privacy loss metric, namely, high variance and values.


Detalles Bibliográficos
2021
machine learning
differential privacy
private aggregation of teacher ensemble
Ciencias Naturales y Exactas
Ciencias de la Computación e Información
Inglés
Agencia Nacional de Investigación e Innovación
REDI
https://hdl.handle.net/20.500.12381/456
https://doi.org/10.3390/make3040039
Acceso abierto
Reconocimiento 4.0 Internacional. (CC BY)
_version_ 1814959255884660736
author Yovine, Sergio
author2 Mayr, Franz
Sosa, Sebastián
Visca, Ramiro
author2_role author
author
author
author_facet Yovine, Sergio
Mayr, Franz
Sosa, Sebastián
Visca, Ramiro
author_role author
bitstream.checksum.fl_str_mv 2d97768b1a25a7df5a347bb58fd2d77f
116d1c86555e21d14da25429e4a9aafb
bitstream.checksumAlgorithm.fl_str_mv MD5
MD5
bitstream.url.fl_str_mv https://redi.anii.org.uy/jspui/bitstream/20.500.12381/456/2/license.txt
https://redi.anii.org.uy/jspui/bitstream/20.500.12381/456/1/make-03-00039.pdf
collection REDI
dc.creator.none.fl_str_mv Yovine, Sergio
Mayr, Franz
Sosa, Sebastián
Visca, Ramiro
dc.date.accessioned.none.fl_str_mv 2021-09-30T13:15:20Z
dc.date.available.none.fl_str_mv 2021-09-30T13:15:20Z
dc.date.issued.none.fl_str_mv 2021-09
dc.description.abstract.none.fl_txt_mv This paper explores the use of Private Aggregation of Teacher Ensembles (PATE) in a setting where students have their own private data that cannot be revealed as is to the ensemble. We propose a privacy model that introduces a local differentially private mechanism to protect student data. We implemented and analyzed it in case studies from security and health domains, and the result of the experiment was twofold. First, this model does not significantly affecs predictive capabilities, and second, it unveiled interesting issues with the so-called data dependency privacy loss metric, namely, high variance and values.
dc.identifier.anii.es.fl_str_mv POS_ICT4V_2016_1_15, FSDA_1_2018_1_154419, FMV_1_2019_1_155913.
dc.identifier.doi.none.fl_str_mv https://doi.org/10.3390/make3040039
dc.identifier.uri.none.fl_str_mv https://hdl.handle.net/20.500.12381/456
dc.language.iso.none.fl_str_mv eng
dc.publisher.es.fl_str_mv MDPI
dc.rights.es.fl_str_mv Acceso abierto
dc.rights.license.none.fl_str_mv Reconocimiento 4.0 Internacional. (CC BY)
dc.rights.none.fl_str_mv info:eu-repo/semantics/openAccess
dc.source.es.fl_str_mv Machine Learning and Knowledge Extraction
dc.source.none.fl_str_mv reponame:REDI
instname:Agencia Nacional de Investigación e Innovación
instacron:Agencia Nacional de Investigación e Innovación
dc.subject.anii.none.fl_str_mv Ciencias Naturales y Exactas
Ciencias de la Computación e Información
dc.subject.es.fl_str_mv machine learning
differential privacy
private aggregation of teacher ensemble
dc.title.none.fl_str_mv An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data
dc.type.es.fl_str_mv Artículo
dc.type.none.fl_str_mv info:eu-repo/semantics/article
dc.type.version.es.fl_str_mv Publicado
dc.type.version.none.fl_str_mv info:eu-repo/semantics/publishedVersion
description This paper explores the use of Private Aggregation of Teacher Ensembles (PATE) in a setting where students have their own private data that cannot be revealed as is to the ensemble. We propose a privacy model that introduces a local differentially private mechanism to protect student data. We implemented and analyzed it in case studies from security and health domains, and the result of the experiment was twofold. First, this model does not significantly affecs predictive capabilities, and second, it unveiled interesting issues with the so-called data dependency privacy loss metric, namely, high variance and values.
eu_rights_str_mv openAccess
format article
id REDI_83becc580b1a9c7ee99448f5f654e059
identifier_str_mv POS_ICT4V_2016_1_15, FSDA_1_2018_1_154419, FMV_1_2019_1_155913.
instacron_str Agencia Nacional de Investigación e Innovación
institution Agencia Nacional de Investigación e Innovación
instname_str Agencia Nacional de Investigación e Innovación
language eng
network_acronym_str REDI
network_name_str REDI
oai_identifier_str oai:redi.anii.org.uy:20.500.12381/456
publishDate 2021
reponame_str REDI
repository.mail.fl_str_mv jmaldini@anii.org.uy
repository.name.fl_str_mv REDI - Agencia Nacional de Investigación e Innovación
repository_id_str 9421
rights_invalid_str_mv Reconocimiento 4.0 Internacional. (CC BY)
Acceso abierto
spelling Reconocimiento 4.0 Internacional. (CC BY)Acceso abiertoinfo:eu-repo/semantics/openAccess2021-09-30T13:15:20Z2021-09-30T13:15:20Z2021-09https://hdl.handle.net/20.500.12381/456POS_ICT4V_2016_1_15, FSDA_1_2018_1_154419, FMV_1_2019_1_155913.https://doi.org/10.3390/make3040039This paper explores the use of Private Aggregation of Teacher Ensembles (PATE) in a setting where students have their own private data that cannot be revealed as is to the ensemble. We propose a privacy model that introduces a local differentially private mechanism to protect student data. We implemented and analyzed it in case studies from security and health domains, and the result of the experiment was twofold. First, this model does not significantly affecs predictive capabilities, and second, it unveiled interesting issues with the so-called data dependency privacy loss metric, namely, high variance and values.engMDPIMachine Learning and Knowledge Extractionreponame:REDIinstname:Agencia Nacional de Investigación e Innovacióninstacron:Agencia Nacional de Investigación e Innovaciónmachine learningdifferential privacyprivate aggregation of teacher ensembleCiencias Naturales y ExactasCiencias de la Computación e InformaciónAn Assessment of the Application of Private Aggregation of Ensemble Models to Sensible DataArtículoPublicadoinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/article//Ciencias Naturales y Exactas/Ciencias de la Computación e Información/Ciencias de la Computación e InformaciónYovine, SergioMayr, FranzSosa, SebastiánVisca, RamiroLICENSElicense.txtlicense.txttext/plain; charset=utf-84746https://redi.anii.org.uy/jspui/bitstream/20.500.12381/456/2/license.txt2d97768b1a25a7df5a347bb58fd2d77fMD52ORIGINALmake-03-00039.pdfmake-03-00039.pdfapplication/pdf1132049https://redi.anii.org.uy/jspui/bitstream/20.500.12381/456/1/make-03-00039.pdf116d1c86555e21d14da25429e4a9aafbMD5120.500.12381/4562021-09-30 10:15:21.98oai:redi.anii.org.uy:20.500.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://www.anii.org.uy/https://redi.anii.org.uy/oai/requestjmaldini@anii.org.uyUruguayopendoar:94212021-09-30T13:15:21REDI - Agencia Nacional de Investigación e Innovaciónfalse
spellingShingle An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data
Yovine, Sergio
machine learning
differential privacy
private aggregation of teacher ensemble
Ciencias Naturales y Exactas
Ciencias de la Computación e Información
status_str publishedVersion
title An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data
title_full An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data
title_fullStr An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data
title_full_unstemmed An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data
title_short An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data
title_sort An Assessment of the Application of Private Aggregation of Ensemble Models to Sensible Data
topic machine learning
differential privacy
private aggregation of teacher ensemble
Ciencias Naturales y Exactas
Ciencias de la Computación e Información
url https://hdl.handle.net/20.500.12381/456
https://doi.org/10.3390/make3040039

Resultados similares