Detecting Complex Money Laundering Patterns with Incremental and Distributed Graph Modeling

Haseeb Tariq; Alen Kaja; Marwan Hassani

cs.LGApr 1, 2026

Detecting Complex Money Laundering Patterns with Incremental and Distributed Graph Modeling

Authors: Haseeb Tariq, Alen Kaja, Marwan Hassani

Organizations: ING Bank Amsterdam, The Netherlands · Eindhoven University of Technology Eindhoven, The Netherlands · Identity & Risk Intelligence Adyen Amsterdam, The Netherlands · Mathematics & Computer Science Eindhoven University of Technology Eindhoven, The Netherlands

Abstract

Money launderers take advantage of limitations in existing detection approaches by hiding their financial footprints in a deceitful manner. They manage this by replicating transaction patterns that the monitoring systems cannot easily distinguish. As a result, criminally gained assets are pushed into legitimate financial channels without drawing attention. Algorithms developed to monitor money flows often struggle with scale and complexity. The difficulty of identifying such activities is further intensified by the (persistent) inability of current solutions to control the excessive number of false positive signals produced by rigid, risk-based rules systems. We propose a framework called ReDiRect (REduce, DIstribute, and RECTify), specifically designed to overcome these challenges. The primary contribution of our work is a novel framing of this problem in an unsupervised setting; where a large transaction graph is fuzzily partitioned into smaller, manageable components to enable fast processing in a distributed manner. In addition, we define a refined evaluation metric that better captures the effectiveness of exposed money laundering patterns. Through comprehensive experimentation, we demonstrate that our framework achieves superior performance compared to existing and state-of-the-art techniques, particularly in terms of efficiency and real-world applicability. For validation, we used the real (open source) Libra dataset and the recently released synthetic datasets by IBM Watson. Our code and datasets are available at https://github.com/mhaseebtariq/redirect.

Figures & tables

Fig. 1 : Conceptualization of the problem formulation for ReDiRect . Left side represents the actual data space; in the next part, nodes are filtered out, either by the primary detection system; or by the output of an earlier ReDiRect run. Finally, the fuzzy communities (per node) are labeled as suspicious by an unsupervised machine learning model.

Anomalies	Count	Threshold	Count / Actual (Threshold Factor)
Actual	600	–	1.00
Top 0.1%	385	T1	0.64
Top 0.2%	770	T2	1.28
Top 0.5%	1926	T3	3.21
Top 1%	3851	T4	6.42

TABLE I : Thresholds used in [ 9 ] to select the top anomalies. The top-x% (predicted anomalies) is calculated from 385,100 accounts.

Dataset	Total	T1 (0.64)	T2 (1.28)	T3 (3.21)	T4 (6.42)
$\mathcal{D}_{lib}^{real}$	600	385	770	1926	3851
$\mathcal{D}_{ibm}^{syn}$ LI	1360	873	1745	4366	8729
$\mathcal{D}_{ibm}^{syn}$ HI	9664	6201	12402	31021	62027

TABLE II : Thresholds used in [ 9 ] for

\mathcal{D}_{ibm}^{syn}

large datasets. LI = Low-Illicit; HI = High-Illicit.

Fig. 2 : Example alerted flow: The yellow are extra (or false-positive); and the red are missing (or false-negative) nodes. The contextual completeness would be → 7 (correctly alerted context) / 11 (alerted with missing context) = 0.64.

Panel of Table 5 — TABLE III : Summary of an example ML (money laundering) flow: transactions, maximum ML share (or contribution), and detection scenarios. Scenario 2 has better context-weighted recall than Scenario 1.

Panel of Figure 6 — Fig. 3 : TPR AUC plots comparing the initial run of ReDiRect (no reduction in data), with the outputs of RM-2 (10% reduction) and RM-3 (12.5% reduction) methods.

	Mean ILT (Investigation Lead Time)
Method		Threshold Factor
	Actual	T1	T2	T3	T4
EgoNetRed [ 9 ]	12.4	102.84	67.24	37.62	23.59
ReDiRect 100%	8.35	23.24	23.49	21.11	17.54
[RM-2] 75%	8.33	23.36	23.4	20.34	16.59
[RM-2] 40%	7.99	21.72	21.4	17.48	13.63

TABLE IV : Mean investigation lead times. Bold font represents the best times. Cells with shaded background represent variations of ReDiRect .

	Alerts					Reports
Method	AUC- T4	TPR- T1	TPR- T2	TPR - T3	TPR - T4	AUC- T4	TPR- T1	TPR- T2	TPR- T3	TPR- T4
EgoNet [ 1 ]	0.599	0.287	0.437	0.660	0.815	0.574	0.146	0.336	0.672	0.850
EgoNetRed [ 9 ]	0.604	0.405	0.514	0.660	0.744	0.603	0.355	0.437	0.696	0.772
ReDiRect 100%	0.617	0.364	0.524	0.660	0.791	0.676	0.318	0.546	0.727	0.955
—"—[RM-2] 75%	0.623	0.364	0.527	0.670	0.804	0.674	0.318	0.546	0.727	0.955
—"—[RM-2] 40%	0.634	0.374	0.529	0.696	0.819	0.653	0.318	0.409	0.727	0.955

TABLE V : Detailed evaluation results on the

\mathcal{D}_{lib}^{real}

dataset. Bold font represents the best version. Red font is when ReDiRect is better than the competitors.

Method	Threshold Factor
Method	T1	T2	T3	T4
[LI] RM-1	0.034	0.075	0.190	0.351
[HI] No reduction	0.050	0.087	0.271	0.450
[HI] RM-1	0.287	0.498	0.675	0.713
[HI] RM-2 50%	0.237	0.477	0.658	0.683

TABLE VI : Context-weighted recall for the

\mathcal{D}_{ibm}^{syn}

datasets. LI = Low-Illicit; HI = High-Illicit.

		p	n	Total
Actual	p’	369	149	518
Actual	n’	3497	63063	66560
Total		3866	63211

TABLE VII : Context-weighted confusion matrix for

\mathcal{D}_{ibm}^{syn}

(high-illicit) [RM-1], with T4 selection threshold.

Fig. 4 : The execution times for each of the heavy duty tasks in the framework, with increasing number of nodes in the $\mathcal{D}_{lib}^{real}$ dataset.

Fig. 5 : Execution times for $\mathcal{D}_{ibm}^{syn}$ large dataset, with increasing level of distribution.

Explore similar work

Extracting Money Laundering Transactions from Quasi-Temporal Graph Representation
Apr 3, 2026Haseeb Tariq, Marwan HassaniFinancial Fraud Detection
ING Bank, Amsterdam, The Netherlands · Eindhoven University of Technology, Eindhoven, The Netherlands
TransXion: A High-Fidelity Graph Benchmark for Realistic Anti-Money Laundering
Apr 19, 2026Keyang Chen, Mingxuan Jiang, Yongsheng Zhao +9Financial Fraud Detection Synthetic Benchmark Generation
Fudan University · Peking University
LineMVGNN: Anti-Money Laundering with Line-Graph-Assisted Multi-View Graph Neural Networks
Mar 24, 2026Chung-Hoo Poon, James Kwok, Calvin Chow +1Financial Fraud Detection Graph Neural Networks
Logistics and Supply Chain MultiTech R&D Centre, Level 11, Cyberport 2, 100 Cyberport Road, Hong Kong · Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong

Apr 3, 2026J/K move · Enter open · S save

Extracting Money Laundering Transactions from Quasi-Temporal Graph Representation

Haseeb Tariq, Marwan Hassani

ING Bank, Amsterdam, The Netherlands · Eindhoven University of Technology, Eindhoven, The Netherlands

Financial Fraud Detection

Open paper AskPDF arXiv

Money laundering presents a persistent challenge for financial institutions worldwide, while criminal organizations constantly evolve their tactics to bypass detection systems. Traditional anti-money laundering approaches mainly rely on predefined risk-based rules, leading to resource-intensive investigations and high numbers of false positive alerts. In order to restrict operational costs from exploding, while billions of transactions are being processed every day, financial institutions are investing in more sophisticated mechanisms to improve existing systems. In this paper, we present ExSTraQt (EXtract Suspicious TRAnsactions from Quasi-Temporal graph representation), an advanced supervised learning approach to detect money laundering (or suspicious) transactions in financial datasets. Our proposed framework excels in performance, when compared to the state-of-the-art AML (Anti Money Laundering) detection models. The key strengths of our framework are sheer simplicity, in terms of design and number of parameters; and scalability, in terms of the computing and memory requirements. We evaluated our framework on transaction-level detection accuracy using a real dataset; and a set of synthetic financial transaction datasets. We consistently achieve an uplift in the F1 score for most datasets, up to 1% for the real dataset; and more than 8% for one of the synthetic datasets. We also claim that our framework could seamlessly complement existing AML detection systems in banks. Our code and datasets are available at https://github.com/mhaseebtariq/exstraqt.

Abstract

Explore similar work

Extracting Money Laundering Transactions from Quasi-Temporal Graph Representation

TransXion: A High-Fidelity Graph Benchmark for Realistic Anti-Money Laundering

LineMVGNN: Anti-Money Laundering with Line-Graph-Assisted Multi-View Graph Neural Networks