Abstract:The important data in the enterprise or organization are stored in the database,so the database often becomes the target of malicious intruders.The traditional firewall can resist the intruders from outside,but it can’t detect the abnormal access from inside.Therefore,this paper proposes an anomaly detection method based on deep feature synthesis and Apriori algorithm,aiming at the shortcomings and defects of database anomaly access detection.Experimental comparison of three other detection algorithms shows that:the new method can extract user features more efficiently,so that the accuracy and efficiency of detection are significantly improved.