Nowadays,　terrorism　has　evolved　into　such　a　destructive　threat　to　the　whole　world　that　it　is　calling　for　an　increasing　devotion　of　professional　researches　and　explorations.　Machine　learning,　as　a　powerful　weapon　to　unveil　the　hidden　knowledge,　has　been　successfully　applied　into　the　anti-terrorism　field.　The　aim　of　this　paper　is　as　follows:　by　implementing　anomaly　detection　algorithm　into　a　famous　terrorism　database　-　GTD,　we　aim　to　locate　the　anomalous　records　within　and　in　doing　so,　we　attempt　to　present　a　list　of　outliers　which　deviate　from　the　rest　of　the　data.　These　finding　anomalous　observations　could　carry　great　hidden　information　which　is　interesting　from　a　terrorism　researcher＇s　perspective.　Then　empirical　analysis　and　experimental　evidence　are　provided　to　support　the　reliability　and　effectiveness　of　the　outcome.　We　present　some　examples　from　the　anomaly　list　and　elaborate　their　abnormality.　Besides,　we　also　validate　the　irregularity　of　these　finding　anomalies　from　the　respect　of　an　improved　classification　precision,　since　these　exceptions　could　be　incurred　by　some　human　errors　which　turn　them　into　noises　and　by　removing　these　noise-like　objects,　we　can　achieve　a　higher　classification　precision.　The　classification　section　is　extended　with　three　sophisticated　classifiers,　Support　Vector　Machine　(SVM),　Naïve　Bayes　(NB)　and　Logistic　Regression　(LR).　©　2017　IEEE.