As network vulnerability gradually attracts people’s attention, detection of critical links (critical links problem) is playing an increasingly significant role in complex networks. According to the features of network society structure, and starting from the division of the network community, we are inspired by the Girvan-Newman (GN) algorithm and propose a critical links detection method on the basis of label propagation algorithm. Aiming at the resources wasting issue caused by assigning labels for each vertex in the algorithm iterative process, and the problem of instable results caused by random iteration, we bind the vertexes that have closer structure together by propagating labels once, and update the labels according to the sequence. Experiments show that the algorithm can figure out the critical links rapidly, stably and efficiently in a complex network.