Evaluating Performance Metrics in Classifying Bitcoin Mixing Services Using Decision Tree Algorithm

Abstract

Bitcoin is a decentralized peer-to-peer (P2P) cryptocurrency system with an innovative payment network. For blockchain applications, ECDSA is formed with public and private keys, especially in Bitcoin, which uses an elliptic curve in a cryptography standard known as Secp256k1 to ensure funds are spent by legitimate owners. Despite the fact that ECDSA is a key component of Bitcoin transactions, today's criminals employ pseudonymous addresses, which make it impossible to track unlawful activity because they don't maintain real-world identities. Bitcoin Fog, Helix Mix, and other mixing services are supposed to provide transaction privacy. However, these services are commonly utilized for concealment, it is more difficult to track criminals. First, using time-frequency analysis, features can be recovered at the network, account, or transaction level to create attributed temporal heterogeneous network Motifs and characterize many forms of address patterns. To increase the performance of evaluation metrics, we introduced a decision-tree machine learning approach for the categorization of bitcoin mixing services from unlabeled addresses in this work. Experiments with bitcoin Kaggle datasets are frequently used to determine the success of our categorization approach.