Analysis of Bitcoin Transactions to Detect Illegal Transactions Using Convolutional Neural Networks

Abstract

The article is devoted to virtual currencies, which is a fast growing and popular market. It was found that for virtual currencies, in particular, for the cryptocurrency Bitcoin, there is a problem of uncontrolled money laundering. This is facilitated by pseudo-anonymization and the presence of illegal exchangers. In this paper, to solve this problem, the method of combining layers in convolutional neural networks is used, which is manifested in the stack layering.In CNN networks, convolutional and erecting layers are usually stacked in a stack, one above the other. The paper proposes a model of Bitcoin transaction analysis to identify anomalies related to money laundering. As such a model, it is proposed to take a combined method, which consists of the method of random forests, enhanced by information from the graph convolutional network, ie, embedded vertices. As a result of the model, we obtained indicators that indicate the presence of possible shadow transactions in the amount of 2-3% of the total market.