The invention discloses a fence intrusion identification method for a fiber fence security protection system. The fence intrusion identification method comprises the steps of: measuring and storing fiber fence vibration signals; intercepting a fiber vibration abnormal signal block greater than threshold value parameters by adopting a zero-crossing rate threshold method; then calculating five groups of characteristic parameters of the fiber vibration abnormal signal block; and finally, training the characteristic parameters by adopting an artificial neural network method, so as to identify an unknown security intrusion vibration signal. According to the fence intrusion identification method, two levels of intrusion behavior identification mechanisms are used in the fiber fence security protection system, namely, a vibration abnormal event is intercepted firstly, and then abnormal event data is subjected to artificial network identification, thereby avoiding calculation of intelligent identification of vibration normal data segments, making the fence security intrusion identification event identification process more targeted, and improving the operation efficiency of a fiber fence intrusion alarm system. In addition, the fence intrusion identification method can effectively reduce the interference of heavy rain and strong winds on the fence security intrusion identification, and distinguish the main intrusion events of the fence precisely.