Location : 基于局部增强和自适应关系聚合的图神经网络...
题目：基于局部增强和自适应关系聚合的图神经网络欺诈检测方法
时间：2025年6月11日（周三）16:20-16:50
地点：中科院数学院南楼N212
报告人：陈进东（北京信息科技大学）
报告摘要：
Fraud detection based on Graph Neural Networks (GNNs) relies on aggregating information from the local neighborhoods, but this process critically fails to address two adversarial tactics: feature and relation camouflage. To enhance the expressive power of GNNs, a fraud detection method based on GNNs with Local Augmentation and Adaptive Relation Aggregation (GNN-LAARA) is proposed. GNN-LAARA includes three key components: 1) a local feature augmentation module based on conditional variational autoencoder (CVAE) generates augmented node representations by learning the conditional distribution of neighboring node features given the central node's features, thereby enhancing the model's ability to identify camouflaged patterns; 2) a dynamic neighbor selector leverages reinforcement learning to filter noisy neighbors via label-aware similarity thresholds; 3) a multi-relational attention aggregator adaptively fuses information from heterogeneous relationships. The effectiveness of GNN-LAARA is validated by two real-world fraud detection datasets. The experimental results validate the effectiveness and generalization of the proposed GNN-LAARA and its components, achieving improvements of up to 2.24% and an average increase of 1.7% on AUC over the state-of-the-art methods. Furthermore, the local feature augmentation module enhances the divergence between normal and abnormal nodes by CVAE, the label-aware similarity metric dynamically prunes relationally camouflaged edges through adaptive thresholds while preserving genuine connectivity, and the multi-relational attention aggregator amplifies the contributions of discriminative edge types for fraud detection.
报告人简介：
陈进东，北京信息科技大学管理科学与工程学院教授，博士生导师。主要研究领域为决策支持系统、机器学习和数据挖掘。曾先后主持国家自然科学基金项目（2项）、国家重点研发计划课题 1项和北京市哲学社会科学基金项目（1项），近年来已经在相关领域国际国内重要期刊上发表相关 SCI/EI/CSSCI 文章10余篇。
CAS,Research Group of Meta-Synthesis and Knowledge Science 京ICP备05002806号-6 文保网安备案号 1101080081 邮箱: mcs@iss.ac.cn
电话：+86 10 82541801