GAME-BC: A Graph Attention Model for Exploring Bitcoin Crime

Abstract

Blockchain technology has been invented as a fundamental technique to the cryptocurrency Bitcoin in 2008, which is decentralized, consensus and cryptographic leger. However, due to the anonymity of the Blockchain, Bitcoin has been becoming one critical finance platform applied to transfer or hidden criminal income by offenders. Bitcoin crime refers to criminal activities which use Bitcoin as a criminal tools, criminal object or criminal settlement. Typical bitcoin crimes include online gambling, money laundering, fraud and more. To address these issues, our work aims to propose an efficient method to find transactions related with Bitcoin crime in the Bitcoin network. Which will efficiently support regulators to combat Bitcoin crimes. Through collecting and concluding kinds of Bitcoin crimes, we find several typical relation patterns among Bitcoin transactions with respect to crimes, and then construct and analysis a bitcoin criminal transaction network. At last, we study a graph neural network model with attention mechanisms to detect illegal transactions. Experimental results show that our method can achieve better classification accuracy, and has the ability to efficiently detect criminal clues and locate related illegal transactions.