Can Edge Addition Be Safe and Effective? Adjacency-Centered Augmentation via Langevin and SDE Diffusion for Self-Supervised Graph Anomaly Detection

Gao, Haokai; Jiang, Menghua; Yang, Xuantao; Liu, Jiale; Li, Yubin; Jiang, Yuncheng

doi:10.24963/ijcai.2026/292

Can Edge Addition Be Safe and Effective? Adjacency-Centered Augmentation via Langevin and SDE Diffusion for Self-Supervised Graph Anomaly Detection

Haokai Gao, Menghua Jiang, Xuantao Yang, Jiale Liu, Yubin Li, Yuncheng Jiang

Download BibTeX Cite

Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence. Main Track. Pages 2626-2634.

https://doi.org/10.24963/ijcai.2026/292

Keywords

Data Mining: Anomaly/outlier detection
Data Mining: Mining graphs
Machine Learning: Self-supervised Learning

Abstract

Edge addition is commonly considered risky in Graph Anomaly Detection (GAD), as random edge addition may induce anomaly–normal connectivity. Consequently, most existing augmentation strategies focus on feature perturbation, edge removal, or subgraph sampling, leaving edge addition largely unexplored. In contrast, we empirically find that moderate and targeted structural completion among normal nodes consistently improves GAD performance, revealing guided edge addition as an overlooked yet effective augmentation dimension. Motivated by this observation, we introduce two adjacency-centered edge generation strategies with complementary mechanisms. One performs a training-free structural completion scheme via spectrum-aware Langevin dynamics, enriching graph connectivity while preserving node features. The other models the joint evolution of node features and graph structure through a stochastic differential equation–based diffusion process, producing structurally coherent and anomaly-aware complementary graphs. Extensive experiments on 12 benchmark datasets with 7 state-of-the-art GAD models demonstrate consistent and substantial improvements in both AUROC and AUPRC. Code and appendices are available at https://github.com/GaoHaokai222/LangGen-JointSDE

Haokai Gao, Menghua Jiang, Xuantao Yang, Jiale Liu, Yubin Li, and Yuncheng Jiang. 2026. Can Edge Addition Be Safe and Effective? Adjacency-Centered Augmentation via Langevin and SDE Diffusion for Self-Supervised Graph Anomaly Detection. In Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26. IJCAI, 2626–2634. https://doi.org/10.24963/ijcai.2026/292

@inproceedings{ijcai2026p292,
  title     = {Can Edge Addition Be Safe and Effective? Adjacency-Centered Augmentation via Langevin and SDE Diffusion for Self-Supervised Graph Anomaly Detection},
  author    = {Gao, Haokai and Jiang, Menghua and Yang, Xuantao and Liu, Jiale and Li, Yubin and Jiang, Yuncheng},
  booktitle = {Proceedings of the Thirty-Fifth International Joint Conference on
               Artificial Intelligence, {IJCAI-26}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {Diego Calvanese},
  pages     = {2626--2634},
  year      = {2026},
  month     = {8},
  note      = {Main Track},
  doi       = {10.24963/ijcai.2026/292},
  url       = {https://doi.org/10.24963/ijcai.2026/292},
}

TY  - CPAPER
TI  - Can Edge Addition Be Safe and Effective? Adjacency-Centered Augmentation via Langevin and SDE Diffusion for Self-Supervised Graph Anomaly Detection
AU  - Gao, Haokai
AU  - Jiang, Menghua
AU  - Yang, Xuantao
AU  - Liu, Jiale
AU  - Li, Yubin
AU  - Jiang, Yuncheng
T2  - Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2026
SP  - 2626
EP  - 2634
DO  - 10.24963/ijcai.2026/292
UR  - https://doi.org/10.24963/ijcai.2026/292
ER  -