Heterogeneous Graph Neural Networks for Malicious Account Detection

TL;DR

This paper introduces GEM, a novel heterogeneous graph neural network designed to detect malicious accounts on Alipay by leveraging account-device graphs and attention mechanisms to improve detection accuracy.

Contribution

The paper presents GEM, the first heterogeneous GNN approach for malicious account detection, incorporating adaptive learning and attention mechanisms for heterogeneous graphs.

Findings

GEM outperforms existing methods in detecting malicious accounts.

The attention mechanism effectively identifies important node types.

The approach demonstrates consistent performance over time.

Abstract

We present, GEM, the first heterogeneous graph neural network approach for detecting malicious accounts at Alipay, one of the world's leading mobile cashless payment platform. Our approach, inspired from a connected subgraph approach, adaptively learns discriminative embeddings from heterogeneous account-device graphs based on two fundamental weaknesses of attackers, i.e. device aggregation and activity aggregation. For the heterogeneous graph consists of various types of nodes, we propose an attention mechanism to learn the importance of different types of nodes, while using the sum operator for modeling the aggregation patterns of nodes in each type. Experiments show that our approaches consistently perform promising results compared with competitive methods over time.