Paper: arXiv 2309.13662
Authors: Haseeb Tariq, Marwan Hassani
Abstract
Money launderers exploit the weaknesses in detection systems by purposefully placing their ill-gotten money into multiple accounts, at different banks. That money is then layered and moved around among mule accounts to obscure the origin and the flow of transactions. Consequently, the money is integrated into the financial system without raising suspicion. Path finding algorithms that aim at tracking suspicious flows of money usually struggle with scale and complexity. Existing community detection techniques also fail to properly capture the time-dependent relationships. This is particularly evident when performing analytics over massive transaction graphs. We propose a framework (called FaSTMAN), adapted for domain-specific constraints, to efficiently construct a temporal graph of sequential transactions. The framework includes a weighting method, using 2nd order graph representation, to quantify the significance of the edges. This method enables us to distribute complex queries on smaller and densely connected networks of flows. Finally, based on those queries, we can effectively identify networks of suspicious flows. We extensively evaluate the scalability and the effectiveness of our framework against two state-of-the-art solutions for detecting suspicious flows of transactions. For a dataset of over 1 Billion transactions from multiple large European banks, the results show a clear superiority of our framework both in efficiency and usefulness.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 8.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced mathematical concepts like higher-order graph representations and temporal graph theory, giving it a moderately high math complexity. It also demonstrates strong empirical rigor through evaluation on a real-world dataset of over 1 billion transactions with comparative benchmarks, making it backtest-ready and data-heavy.
Research Flowchart
flowchart TD A["<b>Research Goal:</b> Detect temporal money laundering flows in billion-scale transactions, overcoming limitations of path-finding & community detection"] --> B["<b>Input Data:</b><br>Over 1 Billion transactions<br>from multiple large European banks"] B --> C["<b>Framework Construction:</b><br>FaSTMAN Framework:<br>Build Temporal Transaction Graph"] C --> D["<b>Methodology:</b><br>2nd Order Graph Representation<br>Weighting Method (Significance Quantification)"] D --> E["<b>Computation:</b><br>Distribute queries on<br>smaller, densely connected networks"] E --> F["<b>Outcomes:</b><br>Identify networks of suspicious flows<br>Superior scalability & efficiency<br>vs. state-of-the-art"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes FaSTMAN, a framework that builds a temporal graph of sequential transactions with a novel weighting method to detect suspicious money flows.
Source sentence
We propose a framework (called FaSTMAN), adapted for domain-specific constraints, to efficiently construct a temporal graph of sequential transactions.
Assumes 5
- Adding the time dimension to the transaction graph could explode the space complexity. model
Source sentence
We show that adding the time dimension to the transaction graph could explode the space complexity.
- Higher‑order (beyond 2nd) graph representations are not considered. model
Source sentence
For our experiments, we did not consider higher (than 2nd) order (3rd, 4th, ..., nth); or multi-order representations.
- The framework can generate the temporal graph in a distributed manner and instantly jump to a point in time. execution
Source sentence
FaSTM∀N enables us to generate the temporal graph in a distributed manner; and to instantly jump to a point in time in the graph.
- FlowScope can produce all flows detected by FaSTM∀N if run multiple times with different k values. other
Source sentence
We will assume that it can produce all the flows detected by FaSTM∀N, after running it multiple times with different values of k.
Show 1 more
- Filtering out the top 0.0001% most highly‑connected accounts does not remove critical suspicious flows. data
Source sentence
We only do that for the top 0.0001% accounts. This results in a reduction of the actual dataset of 1.1 billion transactions, to around 510 million transactions.
Data
- 1.1 billion transactions from 5 large Dutch banks alternative proprietary
Source sentence
Our dataset includes around 1.1 billion transactions from 5 large Dutch banks.
Costs
Not applicable
Finds 2
- FaSTM∀N outperforms two state‑of‑the‑art AML approaches in both scalability and effectiveness. performance
Source sentence
Extensive experimental evaluation with two state-of-the-art AML approaches have proved the superiority of FaSTM∀N in terms of scalability and effectiveness.
- Filtering to the top 0.0001% highly‑connected accounts reduces the dataset from 1.1 billion to about 510 million transactions. other
Source sentence
We only do that for the top 0.0001% accounts. This results in a reduction of the actual dataset of 1.1 billion transactions, to around 510 million transactions.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 10 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.