Illicit Account Detection in the Ethereum Blockchain Using Machine Learning

Abstract

Blockchain is a platform technology for the cryptocurrency's applications like Bitcoin and Ethereum. The purpose of the blockchain is to eliminate the need for third trusted parties such as banks. In recent years and because of the properties of this technology like immutability and transparency, the technology was extended beyond cryptocurrencies and was exploited by various sectors like education, healthcare, finance, energy, government, and IoT providing more privacy, faster transactions and more security. In this research, we investigated Illicit accounts on Ethereum blockchain and proposed a Fraud detection model using three different machine learning algorithms: decision tree (j48), Random Forest and K-nearest neighbors (KNN). These algorithms were applied on a data set obtained from Kaggle.com containing 42 features. We have used the correlation coefficient to select the most effective features and built a new data set using 6 features only. Our research results show a significant improvement in time measurements using the three algorithms and an improvement in the F measure using the Random Forest algorithm.