Cloudcomputingisatechnologythatprovidesuserswithalargestoragespaceandanenormouscomputingpower.However,theoutsourceddataareoftensensitiveandconfidential,andhencemustbeencryptedbeforebeingoutsourced.Consequently,classicalsearchapproacheshavebecomeobsoleteandnewapproachesthatarecompatiblewithencrypteddatahavebecomeanecessity.Forprivacyreasons,mostoftheseapproachesarebasedonthevectormodelwhichisatimeconsumingprocesssincetheentireindexmustbeloadedandexploitedduringthesearchprocessgiventhatthequeryvectormustbecomparedwitheachdocumentvector.Tosolvethisproblem,weproposeanewmethodforconstructingasecureinvertedindexusingtwokeytechniques,homomorphicencryptionandthedummydocumentstechnique.However,1)homomorphicencryptiongeneratesverylargeciphertextswhicharethousandsoftimeslargerthantheircorrespondingplaintexts,and2)thedummydocumentstechniquethatenhancestheindexsecurityproduceslotsoffalsepositivesinthesearchresults.Theproposedapproachexploitstheadvantagesofthesetwotechniquesbyproposingtwomethodscalledthecompressedtableofencryptedscoresandthedoublescoreformula.Moreover,weexploitasecondsecureinvertedindexinordertomanagetheusers'accessrightstothedata.Finally,inordertovalidateourapproach,weperformedanexperimentalstudyusingadatacollectionofonemilliondocuments.Theexperimentsshowthatourapproachismanytimesfasterthananyotherapproachbasedonthevectormodel.
