Explainable Graph Neural Networks for Detecting Coordinated Fraud in Digital Transaction Networks

Coordinated fraud in digital transaction networks involves interconnected accounts that distribute illicit activity across multiple transfers, making suspicious behaviour difficult to identify through isolated transaction analysis. This paper proposes an explainable graph neural network framework for detecting such activity by jointly examining account characteristics, transaction behaviour, and relationships among participants. Digital transactions are represented as a directed graph in which nodes correspond to accounts and edges represent fund transfers, with features capturing transaction amounts, frequency, timing, and account activity. A graph neural network learns neighbourhood representations to estimate account-level fraud risk, while suspicious connections among high-risk accounts are examined to identify candidate fraud groups. An explanation module highlights the account features and transaction subgraphs that contribute most strongly to individual predictions, providing investigators with interpretable evidence for reviewing flagged activity. The proposed evaluation compares the framework with conventional machine learning and graph-based baselines using precision, recall, F1-score, and precision–recall area under the curve. Explanation quality is assessed through fidelity, sparsity, and stability, while chronological data partitioning limits information leakage. The study investigates whether relational learning can improve the identification of coordinated fraudulent behaviour while producing concise and useful explanations. The framework is intended to support transparent fraud screening and human investigation, with particular attention to incomplete coordination labels, changing transaction patterns, and the computational demands of large transaction networks.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-14
DOI
https://doi.org/10.5281/zenodo.24210918
Primary Topic
Advanced Graph Neural Networks
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Explainable Graph Neural Networks for Detecting Coordinated Fraud in Digital Transaction Networks

Anil Kewat, Kismat Chhillar
Zenodo (CERN European Organization for Nuclear Research)
Advanced Graph Neural Networks
article

Explainable Graph Neural Networks for Detecting Coordinated Fraud in Digital Transaction Networks

Anil Kewat, Kismat Chhillar
article en

Abstract

Coordinated fraud in digital transaction networks involves interconnected accounts that distribute illicit activity across multiple transfers, making suspicious behaviour difficult to identify through isolated transaction analysis. This paper proposes an explainable graph neural network framework for detecting such activity by jointly examining account characteristics, transaction behaviour, and relationships among participants. Digital transactions are represented as a directed graph in which nodes correspond to accounts and edges represent fund transfers, with features capturing transaction amounts, frequency, timing, and account activity. A graph neural network learns neighbourhood representations to estimate account-level fraud risk, while suspicious connections among high-risk accounts are examined to identify candidate fraud groups. An explanation module highlights the account features and transaction subgraphs that contribute most strongly to individual predictions, providing investigators with interpretable evidence for reviewing flagged activity. The proposed evaluation compares the framework with conventional machine learning and graph-based baselines using precision, recall, F1-score, and precision–recall area under the curve. Explanation quality is assessed through fidelity, sparsity, and stability, while chronological data partitioning limits information leakage. The study investigates whether relational learning can improve the identification of coordinated fraudulent behaviour while producing concise and useful explanations. The framework is intended to support transparent fraud screening and human investigation, with particular attention to incomplete coordination labels, changing transaction patterns, and the computational demands of large transaction networks.

Zenodo (CERN European Organization for Nuclear Research)
Advanced Graph Neural Networks
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon