The signal source that may be found in perimeter security system of distributed Fiber-optic includes noise signals caused by different environmental state and intrusion signals caused by common kinds of invasion. In order to distinguish noise signals from intrusion signals, a new method based on the M-Z interferometers is proposed. In addition, by using BP neural network, we can recognize different targets while a variety of invasive signals through the Fiber-optic without any ambient noises. Experiment results show that the proposed method is able to differentiate intrusion signals from ambient noises effectively. What’s more, the recognition rate of the system is improved while the false alarm rate is reduced.