The invention provides a fiber intrusion detection method using time-space 2D sparseness of vibration signals to represent K-S test, and aims at a fiber intrusion early-warning system. The method comprises CFAR (Constant False Alarm Rate) detection is carried out on signals of different positions via the CFAR, time-domain K-S test is carried out on data, over-threshold CFAR data is grouped to obtain time-domain to-be-tested data sets, the time interval between data in each time-domain to-be-tested data set is compared with that of prior interference intrusion signals via K-S test, if distribution of the time interval is the same, the detected data set includes interference intrusion signals, and if distribution is different, amplitude K-S test is carried out, namely, the amplitude of data in each time-domain to-be-tested data set is compared with that of the prior interference intrusion signals via K-S test, if distribution of the amplitude is the same, the detected data set is an interference intrusion group, and otherwise, the set is a potential harmful intrusion group. The method detects the intrusion signals in the space-time 2D manner, influence of interference intrusion can rejected, and potential harmful intrusion information is obtained.