It　has　been　widely　recognized　as　a　challenge　to　carry　out　data　analysis　and　meanwhile　preserve　its　privacy　in　the　cloud.　In　this　work,　we　mainly　focus　on　a　well-known　data　analysis　approach　namely　association　rule　mining.　We　found　that　the　data　privacy　in　this　mining　approach　have　not　been　well　considered　so　far.　To　address　this　problem,　we　propose　a　scheme　for　privacy-preserving　association　rule　mining　on　outsourced　cloud　data　which　are　uploaded　from　multiple　parties　in　a　twin-cloud　architecture.　In　particular,　we　mainly　consider　the　scenario　where　the　data　owners　and　miners　have　different　encryption　keys　that　are　kept　secret　from　each　other　and　also　from　the　cloud　server.　Our　scheme　is　constructed　by　a　set　of　well-designed　two-party　secure　computation　algorithms,　which　not　only　preserve　the　data　confidentiality　and　query　privacy　but　also　allow　the　data　owner　to　be　offline　during　the　data　mining.　Compared　with　the　state-of-art　works,　our　scheme　not　only　achieves　higher　level　privacy　but　also　reduces　the　computation　cost　of　data　owners.　©　Springer　International　Publishing　AG,　part　of　Springer　Nature　2018.