同济大学同济大学 数字经济研究中心网络金融安全(国家级)协同创新中心

Financial Risk Control

金融风险防控

研究信用卡欺诈、交易异常与黑产团伙等金融欺诈行为的检测方法。

研究概述

金融风险防控方向,聚焦于金融场景下如信用卡欺诈、交易异常、黑产团伙等欺诈行为的检测,构建涵盖图表示学习、时空建模等多种技术方向的方法体系。探讨如何融合实体属性、行为时序与关系拓扑等不同信号,实现更精准的反欺诈体系。为支撑可复现研究与社区共享,我们开源了统一的欺诈检测研究框架 Antifraud 与模拟半监督金融交易数据集 S-FFSD。2023 年以来,相关成果发表于 KDD、AAAI、IJCAI、WWW 等会议与 TKDE、TPAMI、TIFS 等期刊,研究对象涵盖信用卡与保险欺诈、反洗钱、幌骗与关联交易、信用风险与风险传染,以及反欺诈模型的鲁棒性。

论文成果

  1. Grad 框架
    Grad 框架

    WWW 2025

    Grad: Guided Relation Diffusion Generation for Graph Augmentation in Graph Fraud Detection

    Jie Yang, Rui Zhang, Ziyang Cheng, Dawei Cheng, Guang Yang, Bo Wang

    组织化的欺诈者会模仿平台采集的行为数据,使关键特征与正常用户高度一致,论文称之为「自适应伪装」,这削弱了现有图欺诈检测模型的效果。Grad 以有监督图对比学习扩大欺诈与正常节点之间的差异,并用引导式关系扩散生成器从零生成辅助的同配关系,使微弱的欺诈信号在聚合过程中得到增强、更易被捕获。

    论文arXiv 版本代码摘要
  2. HOGRL 模型框架
    HOGRL 模型框架

    IJCAI 2024

    Effective High-order Graph Representation Learning for Credit Card Fraud Detection

    Yao Zou, Dawei Cheng

    欺诈者常借助多个正常用户进行间接交易以掩盖关联,而图神经网络在深层聚合中的过平滑使模型难以学习这类多跳伪装交易的特征。HOGRL 先构建高阶交易图,直接学习各阶的纯净表示,以识别欺诈者的多跳间接交易;再引入混合专家注意力机制,自动确定不同阶表示的重要性并联合优化检测性能。论文在开源数据集与真实数据集上进行了验证。

    论文arXiv 版本代码摘要
  3. GTAN 模型结构
    GTAN 模型结构

    AAAI 2023

    Semi-supervised Credit Card Fraud Detection via Attribute-Driven Graph Representation

    Sheng Xiang, Mingzhi Zhu, Dawei Cheng, Enxia Li, Ruihui Zhao, Yi Ouyang, Ling Chen, Yefeng Zheng

    真实交易中只有很小比例带有标注,难以充分利用未标注数据中的特征。GTAN 以交易为节点、交易间交互为边构建时序交易图,通过门控时序注意力网络在节点间传递信息、学习交易表示,并借助交易间的风险传播刻画欺诈模式,是一种半监督图神经网络方法。论文在一个真实交易数据集与两个公开欺诈检测数据集上验证,半监督实验考察了仅有少量标注数据时的检测效果。

    论文arXiv 版本代码摘要

20267 篇

  1. IJCAI 2026

    Deep Reinforcement Learning Enhanced Semi-supervised Graph Neural Network for Credit Card Fraud Detection

    论文开放获取全文摘要
  2. IJCAI 2026

    Conspiracy Spoofing Detection via Structure-Augmented Generative Graph Model

    论文开放获取全文摘要
  3. KDD 2026

    Empowering Credit Risk Detection in Weixin Pay with Billion-Scale Deep Graph Learning

    论文arXiv 版本摘要
  4. WWW 2026

    Credit and Power Co-evolution Modeling with Dynamic Graph Learning

    论文摘要
  5. TPAMI 2026IEEE Transactions on Pattern Analysis and Machine Intelligence

    Homophily Edge Augment Graph Neural Network for High-Class Homophily Variance Learning

    论文摘要
  6. TKDE 2026IEEE Transactions on Knowledge and Data Engineering

    Defending Attacks on Anti-Fraud Model With Generative Graph Representations

    论文摘要
  7. AAAI 2026

    Role Perceptual Augmented Temporal Graph Network for Related-party Transaction Detection

    论文开放获取全文代码摘要

20255 篇

  1. TIFS 2025IEEE Transactions on Information Forensics and Security

    Multi-Temporal Partitioned Graph Attention Networks for Financial Fraud Detection

    论文摘要
  2. WWW 2025

    Multi-Granularity Augmented Graph Learning for Spoofing Transaction Detection

    论文摘要
  3. WWW 2025

    Generative Dynamic Graph Representation Learning for Conspiracy Spoofing Detection

    论文arXiv 版本摘要
  4. TKDE 2025IEEE Transactions on Knowledge and Data Engineering

    Mitigating the Tail Effect in Fraud Detection by Community Enhanced Multi-Relation Graph Neural Networks

    论文摘要
  5. TKDE 2025IEEE Transactions on Knowledge and Data Engineering

    Enhancing Attribute-Driven Fraud Detection With Risk-Aware Graph Representation

    论文代码摘要

20243 篇

  1. TKDE 2024IEEE Transactions on Knowledge and Data Engineering

    AdaRisk: Risk-Adaptive Deep Reinforcement Learning for Vulnerable Nodes Detection

    论文代码摘要
  2. IJCAI 2024

    Safeguarding Fraud Detection from Attacks: A Robust Graph Learning Approach

    论文开放获取全文摘要
  3. AAAI 2024

    Pre-trained Online Contrastive Learning for Insurance Fraud Detection

    论文开放获取全文代码摘要

20233 篇

  1. TKDE 2023IEEE Transactions on Knowledge and Data Engineering

    Anti-Money Laundering by Group-Aware Deep Graph Learning

    论文摘要
  2. IJCAI 2023

    Fighting against Organized Fraudsters Using Risk Diffusion-based Parallel Graph Neural Network

    论文开放获取全文摘要
  3. AAAI 2023

    Critical Firms Prediction for Stemming Contagion Risk in Networked-Loans through Graph-Based Deep Reinforcement Learning

    论文开放获取全文摘要

评测结果

Antifraud 仓库公布的检测结果(AUC)
数据集GTANRGTANHOGRLGradSTAGN
YelpChi0.92410.94980.98080.9908—
Amazon0.96300.97500.98000.9789—
S-FFSD0.82860.8461——0.7659

空值表示该方法按仓库说明尚未在该数据集上按同一接口评测,F1 / AP 见原表。