To address the problem of low accuracy in identifying intrusion events in environmental states in the vibration systems, this paper proposes a SCN-based neural network structure for recognizing perimeter intrusion vibration sensor signals. Firstly, four types of perimeter intrusion signals are collected by building a vibration sensor system. Then, the wavelet noise reduction is used to pre-process the signals. After that, the energy features, over-average rate and PAR features of the signals are extracted. Finally, four intrusion events of climbing, touching, impacting and shearing events are recognized by a stochastic configuration network (SCN) neural network. The recognition accuracy of the training set reached 92.7%, and the average recognition accuracy of the test set reached 90.7%. The experimental results show that the proposed method can effectively identify the perimeter intrusion signals.