---
title: "Unsupervised Anomaly Detection in Dynamic Graphs via Compatibility Modeling and Boundary Learning"
authors:
  - "Jiachi Luo"
  - "Shameng Wen"
  - "Ziyang Qiu"
  - "Chen Jiang"
  - "Qi Huang"
  - "Yuxing Tian"
  - "Aiwen Jiang"
doi: "10.24963/ijcai.2026/322"
url: "https://ijcai.org/proceedings/2026/322"
pdf: "https://ijcai.org/proceedings/2026/0322.pdf"
venue: "Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence"
year: 2026
pages: "2897-2905"
track: "Main Track"
keywords:
  - "Data Mining: Anomaly/outlier detection"
  - "Data Mining: Mining graphs"
---

# Unsupervised Anomaly Detection in Dynamic Graphs via Compatibility Modeling and Boundary Learning

Jiachi Luo, Shameng Wen, Ziyang Qiu, Chen Jiang, Qi Huang, Yuxing Tian, Aiwen Jiang

*Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence*, pages 2897-2905. DOI: [10.24963/ijcai.2026/322](https://doi.org/10.24963/ijcai.2026/322)

## Abstract

Anomaly detection in dynamic graphs is essential for monitoring evolving systems such as transaction networks and online platforms. Yet existing methods remain limited in realistic edge-stream settings: snapshot-based approaches discretize continuous interactions and miss fine-grained temporal signals, while many continuous-time models rely on scarce node/edge attributes and often fail to explicitly assess whether a destination is compatible with a source’s recent context. Moreover, under extreme class imbalance and the lack of anomaly labels, unsupervised detectors frequently yield ill-defined normality criteria and ambiguous decision boundaries. We propose BAD, an unsupervised framework for anomaly detection in continuous-time dynamic graphs. BAD adopts a minimalistic design that represents nodes with learnable identity embeddings and performs pairwise compatibility modeling via cross-attention between each destination node and the source’s recent neighbors, enabling direct characterization of context-dependent deviations without requiring attributes. To obtain a principled separating boundary, BAD further integrates normalizing flows to model the distribution of normal interactions and derive likelihood-based anomaly scores. Extensive experiments on four real-world datasets demonstrate that BAD consistently performs well, highlighting its robustness under feature-scarce and label-scarce conditions.