Survey Maps Graph Neural Network Approaches for Detecting Ride‑Hailing Fraud

Global: Survey Maps Graph Neural Network Approaches for Detecting Ride‑Hailing Fraud

A team of six researchers from Sri Lanka released a comprehensive survey on December 29, 2025, examining how graph neural networks (GNNs) can be leveraged to identify fraudulent behavior on ride‑hailing platforms. The paper, titled “A Survey on Graph Neural Networks for Fraud Detection in Ride Hailing Platforms,” was submitted to arXiv and later presented at the 2024 7th International Conference on Artificial Intelligence and Big Data (ICAIBD). Its purpose is to synthesize existing literature, highlight methodological advances, and pinpoint gaps that hinder practical deployment.

Background and Motivation

The authors note that ride‑hailing services have become integral to urban mobility, yet their rapid growth has attracted a range of illicit activities, from fake driver accounts to payment manipulation. Detecting such fraud is complicated by the dynamic, networked nature of transactions, prompting interest in graph‑based machine‑learning techniques.

Fraud Types in Ride‑Hailing

The survey categorizes common fraudulent schemes, including synthetic rider or driver profiles, collusive rating inflation, and location spoofing. By mapping interactions among users, vehicles, and geographic nodes, the authors argue that GNNs are uniquely positioned to capture relational patterns that traditional tabular models miss.

GNN Architectures Reviewed

Among the models surveyed are Graph Convolutional Networks (GCNs), Graph Attention Networks (GATs), and more recent heterogeneous graph transformers. Each architecture is evaluated for its ability to process multimodal data such as trip itineraries, payment histories, and social connections. The authors summarize performance trends reported in prior studies, noting that attention‑based mechanisms often yield higher detection rates on benchmark datasets.

Methodological Challenges

A recurring theme is the severe class imbalance inherent in fraud detection, where legitimate trips vastly outnumber malicious ones. The paper discusses techniques such as oversampling, cost‑sensitive loss functions, and adversarial training to mitigate bias. Additionally, the authors highlight “fraudulent camouflage,” where perpetrators deliberately mimic normal behavior to evade detection, underscoring the need for robust anomaly‑spotting methods.

Research Gaps and Future Directions

The authors identify a shortage of large‑scale, real‑world datasets that reflect the evolving tactics of fraudsters. They call for longitudinal studies that test GNN models in live production environments and for interdisciplinary collaborations that integrate domain expertise from transportation economics and cybersecurity.

Implications for Industry and Academia

By consolidating current knowledge, the survey provides a roadmap for developers seeking to embed GNN‑based fraud detectors into ride‑hailing platforms. It also offers scholars a baseline for comparative experiments, potentially accelerating the translation of academic advances into operational tools.

This report is based on information from arXiv, licensed under Academic Preprint / Open Access. Based on the abstract of the research paper. Full text available via arXiv.