BN-GCN: Detecting Abnormal Bitcoin Transactions Using Graph Neural Networks
Feipeng Jiao, Ruiqing Yue · 2024
With the development and popularity of cryptocurrencies such as Bitcoin, the security and transparency of their transaction systems have received great attention from all sectors of society. The anonymity and decentralised nature of cryptocurrency transactions, while safeguarding user privacy, also provides opportunities for illegal transactions. Traditional abnormal transaction detection methods are often difficult to adapt to the complexity of cryptocurrency transaction networks, thus new techniques are urgently needed to improve detection efficiency and accuracy. The aim of this paper is to propose a Bitcoin anomalous transaction detection model based on graph neural network (BN-GCN model), which introduces a Batch Normalization layer after the GCN convolutional layer to solve the feature magnitude inconsistency, and also prevents overfitting and enhances the generalisation ability through the Dropout layer. The model also contains multiple fully connected layers to fully utilise the extracted features for classification and prediction. Through experiments on Ellipticc++ Dataset, the results show that the proposed graph neural network-based bitcoin anomalous transaction detection model outperforms existing machine learning methods in terms of accuracy and detection speed.