Supervised Machine Learning Techniques in the Bitcoin Transactions. A Case of Ransomware Classification

Abstract

The criminal activities in the bitcoin market have been represented by hacks since the beginnings of cryptocurrency transactions. Ransomware attacks are one of the most dangerous related crimes in the coins market; this type of malware forces victims to pay after encryption of the victim’s file. This paper aims at analysing this context using machine learning techniques through bitcoin transactions data set. The baseline data set containing all the attributes has been compared with two reduced subsets applying feature selection methods in an independent way: Extended Correlation-based Feature Selection (eCFS) and Correlation-based Feature Selection with Exhaustive Search (XCFS). C4.5 algorithm from the trees’ family, PART, which is a rule-based classifier, and kNN, which is lazy algorithm, have been selected to achieve our results. We focus on the kappa and accuracy performance metrics to evaluate on unseen data the trained classification models.