The invention relates to a statistics-based abnormal interface access recognition method, comprising the following steps: step 1, acquiring, by a server, a log that a client accesses the server to form a historical record; step 2, analyzing, by the server, the log that the client accesses the server to obtain a relation graph of access frequencies and the number of users of each interface; step 3, acquiring a relation between access frequencies and the number of the users of a specific interface; step 4, analyzing an abnormal value in the relation between access frequencies and the number of the users; step 5, automatically judging the abnormal access type according to the abnormal value; and step 6, correspondingly processing the abnormal access type. According to the method, statistical analysis is performed on massive client logs, so that user's abnormal use behaviors of illegally accessing the server through the client can be effectively recognized, and an operation support can be provided for operators, wherein the operation support provides a basis for improving products according to the recognized user's abnormal use behaviors in order to eliminate the security risk and shield the user's abnormal use behaviors.