Privacy-Preserving Multidimensional BigData Analytics overBigDataLakes: Models,Techniques,Algorithms

Abstract

It iswellestablishedthathealthinformationprivacyisofcrucialvalueand importancetothegoodexecutionofanalyticalprocesses.Theroleofanalyt- ics, ontheotherhand,isalsoknowntobecriticalforprecisionmedicineas wellasforaccuratehealthcarerecommendations.Whiledataanalyticstryto uncoverpatternsintheunderlyinghealthdatatosupportdecisionmaking, privacy,priorly,ensuresthatthesedataarenotexposingsensitiveinforma- tion abouttheindividualsonwhichtheanalyticalprocessisbeingapplied. Unfortunately,existingresearchworksusuallyputemphasisonsolvingei- ther dataprivacyordataanalyticsissuesinsuchawaythatleadstopoor analytical outcomesintermsofaccuracyfordecisionmakingsupport.This challengemotivatesourneedforajointparadigmbetweendataprivacyand data analyticstoenhancethecapabilitiesofcurrentanalysisofhealthcare data toeventuallyenableprecisionmedicine.Indeed,ajointparadigmwould involvethecreationofalgorithmsthatconsiderafullyintegratedprocessthat enables dataanalyticswhilepreventingthedisclosureofidentityinformation. In thisthesis,weproposeseveralalgorithms,frameworksandtechniquesthat speci callyaddressthepreviousmattersandchallengesindatalakecontexts. Indeed, weaimatdevelopingprivacy-preservingdataanalyticstechniquesin big datalakeenvironmentsbasedondi erentkindsofdatatypesandsettings. In fact,dependingonthegoaloftheprivacypreservinganalyticaltasks,we proposetailoredframeworksthat,weargue,e ectivelyande cientlysupport the creationofhealthrelatedrecommendationsandthus,intheQUALITOP context,supportqualityoflifeaftertreatmentsforcancerpatients.