Article: Efficient blockchain addresses classification through cascading ensemble learning approach Journal: International Journal of Electronic Security and Digital Forensics (IJESDF) 2023 Vol.15 No.2 pp.195 - 210 Abstract: Bitcoin is a pseudonymous, decentralised cryptocurrency that has become one of the most widely utilised digital assets to date. Because of its uncontrolled nature and Bitcoin users' inherent anonymity, it has seen a significant surge in its use for illegal operations. This necessitates the use of unique methods for categorising the addresses of Bitcoin users. This research classifies and predicts the portion of users' activities that are lawful and unlawful on the Bitcoin blockchain. The dataset contains almost 27 billion samples that are divided into nine user acts, five of which were unlawful. To predict cross-validation (CV) accuracy, ensemble learning algorithms are trained and tested. With cross-validation accuracy of 68.63% and 49.64%, respectively, gradient boosting emerged as the best ensemble learning algorithm for classification and prediction, while bagging emerged as the worst. To get the best classification and prediction, hyperparameter tuning is used to find the optimal parameters, which helped to enhance the cross-validation accuracy of the bagging algorithm to 67.70%, with moderate improvements in the rest of the learning algorithms. Inderscience Publishers - linking academia, business and industry through research

Title: Efficient blockchain addresses classification through cascading ensemble learning approach

Authors: Rohit Saxena; Deepak Arora; Vishal Nagar

Addresses: Department of Computer Science and Engineering, Amity School of Engineering and Technology, Amity University, Uttar Pradesh, Lucknow Campus, India ' Department of Computer Science and Engineering, Amity School of Engineering and Technology, Amity University, Uttar Pradesh, Lucknow Campus, India ' Department of Computer Science and Engineering, Pranveer Singh Institute of Technology, Bhauti, Kanpur, India

Abstract: Bitcoin is a pseudonymous, decentralised cryptocurrency that has become one of the most widely utilised digital assets to date. Because of its uncontrolled nature and Bitcoin users' inherent anonymity, it has seen a significant surge in its use for illegal operations. This necessitates the use of unique methods for categorising the addresses of Bitcoin users. This research classifies and predicts the portion of users' activities that are lawful and unlawful on the Bitcoin blockchain. The dataset contains almost 27 billion samples that are divided into nine user acts, five of which were unlawful. To predict cross-validation (CV) accuracy, ensemble learning algorithms are trained and tested. With cross-validation accuracy of 68.63% and 49.64%, respectively, gradient boosting emerged as the best ensemble learning algorithm for classification and prediction, while bagging emerged as the worst. To get the best classification and prediction, hyperparameter tuning is used to find the optimal parameters, which helped to enhance the cross-validation accuracy of the bagging algorithm to 67.70%, with moderate improvements in the rest of the learning algorithms.

Keywords: blockchain; Bitcoin; ensemble learning; machine learning; ML; classification; anonymity.

DOI: 10.1504/IJESDF.2023.129278

International Journal of Electronic Security and Digital Forensics, 2023 Vol.15 No.2, pp.195 - 210

Received: 01 Mar 2022
Accepted: 22 Jun 2022
Published online: 02 Mar 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Efficient blockchain addresses classification through cascading ensemble learning approach

Keep up-to-date