Parallelizing record linkage for disclosure risk assessment

Joan Guisado-Gámez, Arnau Prat-Pérez, Jordi Nin, Victor Muntés-Mulero, Josep Ll Larriba-Pey

Producció científica: Capítol de llibreContribució a congrés/conferènciaAvaluat per experts

4 Cites (Scopus)

Resum

Handling very large volumes of confidential data is becoming a common practice in many organizations such as statistical agencies. This calls for the use of protection methods that have to be validated in terms of the quality they provide. With the use of Record Linkage (RL) it is possible to compute the disclosure risk, which gives a measure of the quality of a data protection method. However, the RL methods proposed in the literature are computationally costly, which poses difficulties when frequent RL processes have to be executed on large data. Here, we propose a distributed computing technique to improve the performance of a RL process. We show that our technique not only improves the computing time of a RL process significantly, but it is also scalable in a distributed environment. Also, we show that distributed computation can be complemented with SMP based parallelization in each node increasing the final speedup.

Idioma originalAnglès
Títol de la publicacióPrivacy in Statistical Databases - UNESCO Chair in Data Privacy International Conference, PSD 2008, Proceedings
EditorSpringer Verlag
Pàgines190-202
Nombre de pàgines13
ISBN (imprès)3540874704, 9783540874706
DOIs
Estat de la publicacióPublicada - 2008
Publicat externament
EsdevenimentInternational Conference on Privacy in Statistical Databases, PSD 2008 - Istanbul, Turkey
Durada: 24 de set. 200826 de set. 2008

Sèrie de publicacions

NomLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volum5262 LNCS
ISSN (imprès)0302-9743
ISSN (electrònic)1611-3349

Conferència

ConferènciaInternational Conference on Privacy in Statistical Databases, PSD 2008
País/TerritoriTurkey
CiutatIstanbul
Període24/09/0826/09/08

Fingerprint

Navegar pels temes de recerca de 'Parallelizing record linkage for disclosure risk assessment'. Junts formen un fingerprint únic.

Com citar-ho