Spam　filtering　has　witnessed　a　booming　interest　in　the　recent　years,　due　to　the　increased　abuse　of　email.　This　paper　presents　SpamTerminator,　a　personal　anti-spam　filtering　add-in　of　Outlook.　Outstanding　characteristics　of　SpamTerminator　are　as　follows.　First,　it　provides　eleven　filters　including　rule-based,　white　lists,　black　lists,　four　single　filters,　and　four　ensemble　filters.　As　a　result,　SpamTerminator　can　automatically　work　for　users　in　different　stages　even　if　they　do　not　train　machine　learning-based　filters.　Secondly,　by　using　our　proposed　method　named　TPL　(Two-Phases　Learning)　to　combine　multiple　disparate　classifiers,　ensemble　filters　can　achieve　excellent　discrimination　between　spam　and　legitimate　mail.　©　2009　Springer-Verlag.