With the development of the Internet, there has been an proliferation of coordinated groups employed to manipulate and steer public opinion, rendering their detection and analysis increasingly vital. In recent years, coordination detection have improved detection accuracy by constructing multi-view graph to capture richer structural features. However, existing multi-view graph coordination detection methods overlook the semantic features of users across different behaviors and struggle to extract coordinated features of diverse behaviors through general methods, leading to high subjectivity and unstable detection results. To address these issues, this paper proposes a sampling-enhanced multi-view graph contrastive learning for coordination detection(Mv-GCD). this method utilizes graph neural networks(GNN) to capture and integrate semantic and structural features across multiple behaviors. It effectively captures the fundamental temporal and structural features of coordinated behaviors through a node sampling strategy. Subsequently, the model undergoes training through node-level and graph-level contrastive learning to obtain node vector representations, which are used to differentiate coordinated and non-coordinated groups via clustering algorithm. Experimental results demonstrate that Mv-GCD outperforms existing methods. Ablation experiments further explore the impact of different parameters on the method.

The (( k,p ))-core model was recently proposed to capture engagement dynamics by considering both intra-community interactions (i.e., the k -core structure) and inter-community interactions (i.e., the p -fraction property). It is a refinement of the classic k -core, by introducing an extra parameter p to customize the engagement within a community at a finer granularity. In this paper, we study the problem of maintaining all (k,p)-cores (essentially, maintaining the p-numbers for all vertices) for dynamic graphs. The existing Global approach conducts a global peeling, almost from scratch, for all vertices whose old p-numbers are within a computed range [p - ,p + ], and thus is inefficient. We propose a new Local approach which conducts local searches starting from the two end-points of the newly inserted or deleted edge, and then iteratively expands the search frontier by including their neighbors. Our algorithm is designed based on several fundamental properties that we prove in this paper to characterize the necessary condition for a vertex's p-number to change. Compared to Global, our Local approach implicitly obtains the optimal affected p-number range [p - * ,p + * ] ⊆ [p - ,p + ], and further skips many vertices whose p-numbers are within this range. Experimental results show that Local is on average two orders of magnitude faster than Global.

In social media, coordinated groups spread false information and guide public opinion through organized social behaviors. Coordination detection methods based on behavioral sequences identify coordinated groups among users by extracting temporal and structural data. However, existing methods generally concentrate on identification and analysis from an individual perspective, lacking evaluation and detection from a group perspective to measure the degree of coordination, which results in low accuracy of detection results. To address these issues, we propose an unsupervised evaluation method of Relative Coordination Degree of Groups (RCDG). This method models the global time and differentiated network features from group perspective using Neural Time Point Process (NTPP) and Graph Contrastive Learning (GCL), and evaluates the relative degree of coordination based on the fused features. Experimental results show that RCDG can effectively evaluate the relative degree of coordination among groups. Furthermore, it can achieve an accuracy rate of up to 99.8% in detecting coordinated groups.