Leveraging Graph Convolutional Networks for the Detection of Illicit Bitcoin Transactions

Kannekanti, Kavitha (2024) Leveraging Graph Convolutional Networks for the Detection of Illicit Bitcoin Transactions. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (955kB) | Preview

Abstract

With the advancement of cryptocurrencies, especially Bitcoin, the rate and instances of crimes have increased to become a challenge to the agencies responsible for regulation. The heuristic techniques, which are commonly used in detecting frauds are traditional methods, take a lot of time and cannot be easily scaled. This research proposes a new approach called Graph-based Transaction Anomaly Detection (GTAD) that employs Graph Convolutional Networks (GCNs), Temporal Graph Convolutional Networks (T-GCNs) and transformers to enhance the identification of Illicit Bitcoin transactions. For the modelling of Bitcoin transaction networks, GTAD builds a directed graph where temporal dynamics are associated with time-based snapshots. For feature learning, this approach employs GCNs to extract meaningful node embeddings, while temporal attention helps identify important temporal and spatial patterns in transaction data. Also, the hierarchical attention mechanism gives preference to massive transactions and concentrates on areas of high fraud risk. The proposed model is trained under supervised learning paradigm, with a weighted cross entropy loss function to handle class imbalance. Comparing GTAD with various machine learning methods for detecting illicit activities shows that GTAD provides higher accuracy and better scalability. The performance of the GTAD is higher than that of other models with 98.49% accuracy and 0.94 macro F1-score. These findings demonstrate the applicability of the model to prevent financial crimes in the cryptocurrency context and can help improve the detection of fraud in the constantly developing digital currency environment.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Hava Muntean, Cristina UNSPECIFIED
Uncontrolled Keywords:	Fraudulent Transactions; Graph-based Transaction Anomaly Detection (GTAD); Graph Convolutional Networks (GCNs); Bitcoin transactions
Subjects:	H Social Sciences > HV Social pathology. Social and public welfare > Criminology > Crimes and Offences Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science H Social Sciences > HG Finance > Money > Digital currency > Cryptocurrencies Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	02 Sep 2025 15:21
Last Modified:	02 Sep 2025 15:21
URI:	https://norma.ncirl.ie/id/eprint/8722

Actions (login required)

View Item