The present disclosure relates to a method performed by a network node (54) in a communication network (50) comprising a plurality of communication devices (51), for finding temporally connected connection patterns of the communication devices in the network. The method comprises identifying signalling between the communication devices during a predefined time duration to form a main communication graph in which the plurality of communication devices are nodes. The method also comprises partitioning the main communication graph for a time period comprised in said time duration to capture temporally connected signalling between some of the communication devices as illustrated in the main communication graph in a partitioned communication graph. The method also comprises forming at least one subgraph from the partitioned communication graph, comprising m nodes corresponding to m communication devices of said some of the communication devices. The method also comprises constructing a binary vector of the subgraph over the time period divided into a plurality of sampling periods corresponding to components in the vector such that, for each component in the vector, the component value is set to “1” if, in accordance with the subgraph, signalling occurs between at least two of the m communication devices during the sampling period corresponding to the element, and the component value is otherwise set to “0”. The method also comprises testing whether the signalling between the different m communication devices used for constructing the binary vector is temporally connected by checking that the number of consecutive “0” element(s), if any, between any two “1” elements is less than a predefined ΔT.