TransXion：面向真实反洗钱的高保真图基准
摘要
洗钱活动构成全球金融系统的严重风险，推动了机器学习在交易监控中的广泛采用。然而，由于缺乏真实可靠的基准，进展受阻。现有交易图数据集存在两个普遍局限：(i) 仅提供匿名标识符之外的稀疏节点级语义，(ii) 依赖模板驱动的异常注入，导致基准偏向静态结构构式，评估模型鲁棒性显得过于乐观。我们提出TransXion，一个为反洗钱（AML）研究构建的基准生态系统，集成正常活动的轮廓感知仿真与非模板式非法子图的随机合成。TransXion联合建模持久实体轮廓与条件交易行为，使能够评估"超出角色"异常，即观察到的活动与实体的社会经济背景相矛盾。 resulting dataset comprises approximately 3 million transactions among 50,000 entities, each endowed with rich demographic and behavioral attributes. Empirical analyses show that TransXion reproduces key structural properties of payment networks, including heavy-tailed activity distributions and localized subgraph structure. Across a diverse array of detection models spanning multiple algorithmic paradigms, TransXion yields substantially lower detection performance than widely used benchmarks, demonstrating increased difficulty and realism. TransXion provides a more faithful testbed for developing context-aware and robust AML detection methods. The dataset and code are publicly available at https://github.com/chaos-max/TransXion。
引用
@article{arxiv.2604.17420,
title = {TransXion: A High-Fidelity Graph Benchmark for Realistic Anti-Money Laundering},
author = {Keyang Chen and Mingxuan Jiang and Yongsheng Zhao and Zeping Li and Zaiyuan Chen and Weiqi Luo and Zhixin Li and Sen Liu and Yinan Jing and Guangnan Ye and Xihong Wu and Hongfeng Chai},
journal= {arXiv preprint arXiv:2604.17420},
year = {2026}
}