🤖 AI Summary
This paper addresses the detection and localization of anomalous cliques—fully connected subgraphs—in heterogeneous networks. Existing methods lack theoretical guarantees and computational feasibility beyond homogeneous models such as Erdős–Rényi graphs. To overcome this, we propose a statistically principled framework based on ego-nets—the first-order neighborhoods of nodes—as unified, parallelizable primitives. Our approach integrates permutation testing with distributed computation, enabling adaptation to diverse heterogeneous models, including stochastic block models (SBMs) and degree-corrected variants. The method is model-agnostic, computationally efficient, and provides rigorous statistical interpretability. Extensive experiments on synthetic benchmarks and real-world networks—including Karate and Dolphin—demonstrate substantial improvements in detection accuracy and robustness against structural heterogeneity. These results validate the framework’s generality and practical utility for clique-based anomaly detection in complex, non-uniform networked systems.
📝 Abstract
Network data has rapidly emerged as an important and active area of statistical methodology. In this paper we consider the problem of anomaly detection in networks. Given a large background network, we seek to detect whether there is a small anomalous subgraph present in the network, and if such a subgraph is present, which nodes constitute the subgraph. We propose an inferential tool based on egonets to answer this question. The proposed method is computationally efficient and naturally amenable to parallel computing, and easily extends to a wide variety of network models. We demonstrate through simulation studies that the egonet method works well under a wide variety of network models. We obtain some fascinating empirical results by applying the egonet method on several well-studied benchmark datasets.
