With　the　development　of　image　processing　technology　and　the　pervasiveness　of　mobile　devices,　face　recognition,　which　can　be　used　to　offer　convenient　and　efficient　individual　authentication　service,　has　attracted　considerable　interest　in　recent　years.　However,　people＇s　concern　about　their　face　data　being　leaked　during　the　face　recognition　process　impedes　the　flourish　of　face　recognition.　To　address　this　problem,　we　present　a　novel　privacy-preserving　online　face　recognition　scheme　over　encrypted　outsourced　data,　named　EPFR.　With　EPFR,　a　user　can　achieve　secure,　accurate　and　efficient　authentication　service　without　disclosing　her/his　face　data.　Specifically,　an　improved　homomorphic　encryption　technology　is　introduced　to　provide　an　efficient　online　face　recognition　service　based　on　the　Eigenface　algorithm.　Through　extensive　analysis,　we　show　that　users＇　face　data　are　kept　confidential　during　the　online　face　recognition　process.　In　addition,　we　implement　the　scheme　with　a　real　face　database,　and　simulation　results　demonstrate　that　the　scheme　can　be　used　to　provide　efficient　and　accurate　online　face　recognition　service.　©　2018　IEEE.