ShareFlow: An Efficient Framework for Multi-Query Continuous Subgraph Matching

In this work, we study the Multi-Query Continuous Subgraph Matching (MQCSM) problem in dynamic graphs, which is widely used in various real-world applications, e.g., fraud detection, anomaly detection, and social network analysis. Existing work on Continuous Subgraph Matching (CSM) is either single-query-oriented, too resource-intensive to be applied to multiple simultaneous queries, or multi-query-oriented with significant limitations, including ineffective indices, repeated search, etc. Motivated by them, we propose an efficient MQCSM framework ShareFlow in this work. It first introduces an annotated graph structure that merges multiple queries into a compact representation and enables shared computations. Then, a twolevel indexing mechanism is devised in ShareFlow to accelerate MQCSM processing. In particular, it includes a shared candidate graph (SCG) and an effective candidate graph (ECG) that reduce search space and improve backtracking efficiency. Next, a block-based search (BBS) engine is designed in ShareFlow to avoid repeated search paths in the Cartesian products of matches of sub-blocks. Experimental evaluations on real datasets confirm that ShareFlow offers substantial performance benefits over existing algorithms. Its components collectively contribute to a more efficient and scalable approach to MQCSM, outperforming NewSP, RapidFlow, CaLiG, Symbi, and MQ-Match by up to 18.85×, 14.27×, 457.74×, 172.03 ×, and 202.45×, respectively.