The public security system is increasingly utilizing technology and big data for criminal network identification, making the identification of criminal suspects through social data a crucial aspect of current criminal network identification. Based on our research into the verbal features of networks used to portray criminal suspects, we constructed a criminal network identification system using social networks and complex network theory. We then designed the overall framework and main functional modules. We established the case data warehouse of public security agencies and built the social relationship network of criminal suspects. This paper compares the effectiveness of the complex network (CN) algorithm with other community delineation algorithms and utilizes the criminal network identification system it constructs for case data mining and analysis. The extended modularity values of the CN algorithm proposed in this paper on four real datasets are the largest among all algorithms, and the community delineation effect is optimal. In the mining of criminal gangs, a1, a3, and a15 represent the heads of the gangs, while a56 shares similar duty content with a1, a3, and a15, albeit at a lower rank. A65 is in charge of connecting and communicating with gang members.