Secure Multiparty Computation for Privacy-Preserving Data Mining

《Secure Multiparty Computation for Privacy-Preserving Data Mining》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、SecureMultipartyComputationforPrivacy-PreservingDataMiningYehudaLindellBennyPinkasyMay6,2008AbstractInthispaper,wesurveythebasicparadigmsandnotionsofsecuremultipartycomputationanddiscusstheirrelevancetotheeldofprivacy-preservingdatamining.Inadditio

2、ntore-viewingdenitionsandconstructionsforsecuremultipartycomputation,wediscusstheissueofeciencyanddemonstratethedicultiesinvolvedinconstructinghighlyecientprotocols.Wealsopresentcommonerrorsthatareprevalentintheliteraturewhensecuremultipartycompu

3、tationtechniquesareappliedtoprivacy-preservingdatamining.Finally,wediscusstherelationshipbetweensecuremultipartycomputationandprivacy-preservingdatamining,andshowwhichproblemsitsolvesandwhichproblemsitdoesnot.1IntroductionBackground.Privacy-preservin

4、gdataminingconsiderstheproblemofrunningdataminingalgorithmsoncondentialdatathatisnotsupposedtoberevealedeventothepartyrunningthealgorithm.Therearetwoclassicsettingsforprivacy-preservingdatamining(althoughthesearebynomeanstheonlyones).Intherst,theda

5、taisdividedamongsttwoormoredierentparties,andtheaimistorunadataminingalgorithmontheunionoftheparties'databaseswithoutallowinganypartytoviewanyoneelse'sprivatedata.Inthesecond,somestatisticaldatathatistobereleased(sothatitcanbeusedforresearchusingsta

6、tisticsand/ordatamining)maycontaincondentialdataandsoisrstmodiedsothat(a)thedatadoesnotcompromiseanyone'sprivacy,and(b)itisstillpossibletoobtainmeaningfulresultsbyrunningdataminingalgorithmsonthemodieddataset.Inthispaper,wewillmainlyrefertoscenar

7、iosofthersttype.Aclassicalexampleofaprivacy-preservingdataminingproblemofthersttypeisfromtheeldofmedicalresearch.Considerthecasethatanumberofdierenthospitalswishtojointlyminetheirpatientdataforthepurposeofmedicalresearch.Furthermore,letusassumeth

8、atprivacypolicyandlawpreventsthesehospitalsfromeverpoolingtheirdataorrevealingittoeachother,duetothecondentialityofpatientrecords.Insuchacase,classicaldataminingsolutionscannotbeused.Ratheritisnecessarytondasolutionthatenablesthehospitalstocomputet