Identification of suspicious transactions in the bitcoin network: analysis, features, and machine learning algorithms

Abstract

This article talks about one of the most popular cryptocurrencies today bitcoin. The principles of bitcoin and blockchain technologies are considered, where the proc and cons of bitcoin are listed. The focus of the research work is on the analysis of the use of cryptocurrencies, namely bitcoin for illegal and criminal purposes. Particular attention is paid to identifying a set of signs to identify suspicious activity in the bitcoin network. During research work reviewed and analyzed many scientific publications and articles, and from these studies, signs were identified that are key attributes when evaluating a Bitcoin transaction for suspicious or questionable transactions. Based on these features, an input dataset of 15 attributes with about 100,000 transactions was generated for further use in building a model that will identify suspicious bitcoin transactions. Using this dataset, models were created and trained to detect suspicious transactions in the bitcoin network based on several machine learning algorithms, such as random forest, logistic regression, k-nearest neighbors, decision tree. The article also provides comparisons of the results of machine learning algorithms, and in addition, selects the best algorithm that showed the best accuracy.