A Dataset-Driven Validation of Structural Constraints in Network Traffic–Based Detection of Financial Crimes

Modern financial ecosystems are increasingly exposed to money laundering, fraud, phishing, and transaction obfuscation within encrypted and decentralized network environments. While prior research identifies detection challenges conceptually, limited work empirically validates how structural constraints manifest across real-world network datasets. This study presents a dataset-driven validation of structural constraints in network traffic–based detection of financial crime using three progressively scaled real-time datasets: NetTran3, NetTran4, and NetTran5. Quantitative analysis shows that encrypted traffic obscures 22.03% of transactions in NetTran3, rises to 39.29% in NetTran4, and declines to 18.24% in NetTran5, reflecting evolving encryption dynamics. Tumbling services—used for transaction obfuscation—are observed in 34.42%, 34.01%, and 40.22% of transactions, respectively, indicating increasing transactional fragmentation. Additional constraints include scalability pressure, cross-chain complexity, adversarial manipulation, and data imbalance. Rather than benchmarking detection algorithms, this study systematically evaluates how these infrastructural characteristics intensify detection difficulty as dataset size and complexity increase. The findings provide empirical grounding for literature-identified constraints and establish a structured foundation for advancing robust, scalable, and privacy-aware financial crime detection systems.

Financial Crime Detection, Network Traffic Analysis

1. B. Liu, Z. Zhang, X. Xu, and M. Luo, “A comprehensive study on anomaly detection in financial systems,” J. Netw. Comput. Appl., vol. 174, p. 102695, 2021. DOI: 10.1016/j.jnca.2020.102695. [Google Scholar] [Crossref]

2. A. Venčkauskas, R. Maskeliūnas, and R. Damaševičius, “Cryptographic methods in anomaly detection for blockchain systems,” Future Gener. Comput. Syst., vol. 134, p. 203214, 2024. DOI:10.1016/j.future.2023.103214. [Google Scholar] [Crossref]

3. M. Turner, A. Rigby, and T. Green, “Addressing scalability issues in fraud detection: A graph-based approach,” Pattern Recognit., vol. 115, p. 107968, 2020. DOI: 10.1016/j.patcog.2020.107968. [Google Scholar] [Crossref]

4. A. Jeyakumar, D. Nathan, and T. Sam, “Machine learning methods for detecting cryptocurrency-related financial crimes,” Comput. Secur., vol. 109, p. 102212, 2021. DOI: 10.1016/j.cose.2020.102212. [Google Scholar] [Crossref]

5. J. Kang and Q. Buu, “Privacy-preserving techniques in anomaly detection models,” J. Comput. Sci., vol. 147, p. 207345, 2024. DOI: 10.1016/j.jocs.2023.103456. [Google Scholar] [Crossref]

6. R. Pocher, F. Lemos, and K. Gupta, “Challenges in regulatory frameworks for cryptocurrency transactions,” Int. J. Finance, vol. 136, p. 409823, 2023. DOI: 10.1016/j.ijfin.2022.104823. [Google Scholar] [Crossref]

7. S. Nurmara, C. Lee, and D. Wong, “Advances in anomaly detection using ensemble learning techniques,” Inf. Syst., vol. 153, p. 103456, 2023. DOI: 10.1016/j.is.2022.102345. [Google Scholar] [Crossref]

8. F. N. Mauliddiah and W. Sun, “Implementing graph neural networks for detecting financial crimes,” Graph Theory Appl., vol. 12, p. 213245, 2023. DOI: 10.1016/j.gta.2023.123456. [Google Scholar] [Crossref]

9. C. Dumitrescu, R. Martin, and S. Andrei, “Novel approaches to address scalability in transaction analysis,” Distrib. Ledger Res., vol. 94, p. 211345, 2022. DOI: 10.1016/j.dlr.2022.102345. [Google Scholar] [Crossref]

10. D. Gabrielli, P. Ochoa, and H. Nguyen, “Federated learning applications in anomaly detection,” Artif. Intell. Rev., vol. 184, p.20385, 2024. DOI: 10.1016/j.airev.2023.103456. [Google Scholar] [Crossref]

12. A. Khan, F. Raza, and M. Qureshi, Privacy and scalability in blockchain-based financial systems,” Digit. Ledger J., vol. 120, p. 103456, 2023. DOI: 10.1016/j.dlj.2022.102345. [Google Scholar] [Crossref]

13. P. Ghosh, Z. Anwar, and T. Hussain, “Addressing imbalanced datasets in financial fraud detection,” Mach. Learn. Secur. Appl., vol. 45, p. 102345, 2023. DOI: 10.1016/j.mlsa.2022.102345. [Google Scholar] [Crossref]

14. https://drive.google.com/file/d/1qLVeSInr uY6a3-W5bm3ygTRFSnRJ3MY/view?usp=drive_link. [Google Scholar] [Crossref]

15. https://drive.google.com/file/d/1GdS3EcG pIUwKvkdOxJcGinIUwsby1LAO/view?u sp=drive_link. [Google Scholar] [Crossref]

16. https://drive.google.com/file/d/18nsR1g66 hQG3KxCNORtdfXFeiVHdguSC/view?u sp=drive_link. [Google Scholar] [Crossref]

17. Muhammad Nuraddeen Ado. (2025). Network Transaction Datasets [Data set]. Kaggle. https://doi.org/10.34740/KAGGLE/DSV/1 1300360 [Google Scholar] [Crossref]

• What the Desert Fathers Teach Data Scientists: Ancient Ascetic Principles for Ethical Machine-Learning Practice
• Comparative Analysis of Some Machine Learning Algorithms for the Classification of Ransomware
• Comparative Performance Analysis of Some Priority Queue Variants in Dijkstra’s Algorithm
• Transfer Learning in Detecting E-Assessment Malpractice from a Proctored Video Recordings.
• Dual-Modal Detection of Parkinson’s Disease: A Clinical Framework and Deep Learning Approach Using NeuroParkNet