Privacypreservingdataminingalgorithmsarecrucialforthepersonaldataanalysis,suchasmedicalandfinancialrecords.Thispaperfocusesonfeatureselectionandproposesanewprivacypreservingdistributedalgorithm,whichcaneffectivelyselectfeaturesbasedondifferentialprivacyandGiniindexundertheMapReduceframework.Atthesametime,thetheoreticanalysisforprivacyguaranteeisalsopresented.Someexperimentsareconductedonbench-markdatasets,thesimulationresultsindicatethatduringtheselectionofimportantfeatures,theproposedalgorithmcanpreserveprivacyinformationtoacertainextentwithlesstimecostthanoncentralizedcounterpart.