Comparative Analysis of Machine Learning Based Fraud Detection Techniques in Blockchain
Volume 18, Issue 1, 2024
Download| Author(s): |
Sabih Hida Tahir SZABIST University, Karachi Campus , sabihhida@hotmail.com Muhammad Raza SZABIST University, Gharo Campus, Pakistan, dr.raza@ghr.szabist.edu.pk Rageshwari Haryani SZABIST University, Karachi Campus, Pakistan, rageshwariharyani1997@gmail.com Almina Sherish SZABIST University, Karachi Campus, Pakistan, alminasehrish88@gmail.com |
|---|---|
| Abstract | Fraudulent activities like phishing and pump-dump schemes clearly threaten the integrity and reliability of decentralized platforms, especially Ethereum. This paper compares the quality of fraud detection methods in Ethereum's platform. It emphasizes the potential of unsupervised and supervised learning algorithms applied to Ethereum. The aim is to have an advanced system capable of firmly protecting Ethereum by detecting fraud and putting a stop to it. This paper collected transactional data on Ethereum, smart contract interactions, and past fraud activities possibly significant to net miners. It further proposed fine-grained features targeting Ethereum transaction nuances, which are important early signs of fraud. The paper takes an integrated approach in comparing traditional supervised methods such as Random Forest, eXtreme Gradient Boosting, Decision Trees, K-Nearest Neighbors (KNN), Support Vector Machines (SVM), Naive Bayes, Logistic Regression, and Linear Discriminant Analysis (LDA), versus unsupervised learning like outlier detection or clustering algorithms such as K-Means, Gaussian Mixture Models (GMM), and BIRCH. Unlike most Ethereum fraud detection studies that rely heavily on supervised techniques, this one highlights the lack of unsupervised techniques and shifts the spotlight to a comparative analysis with three unsupervised algorithms. In addition, this paper also compares the algorithms on time efficiency. Benchmarking with traditional supervised techniques indicates that unsupervised learning is more effective in detecting new fraudulent patterns. Overall evaluation was further broken down under headings of precision, recall, F1-score and Silhouette score. It proposes a proactive fraud prevention system for Ethereum, having foreseen an event before it actually happens. The goal is to maintain security in Ethereum, as well as other decentralized networks, providing flexible defenses against this rapidly evolving form of crime. |
| Keywords | Blockchain Security, Ethereum Network, Fraud Detection, Unsupervised Learning, Comparative Analysis, Proactive Security, Anomaly Detection, Cryptocurrency |
| Year | 2024 |
| Volume | 18 |
| Issue | 1 |
| Type | Research paper, manuscript, article |
| Journal Name | Journal of Information & Communication Technology | Publisher Name | ILMA University | Jel Classification | - | DOI | - | ISSN no (E, Electronic) | 2075-7239 | ISSN no (P, Print) | 2415-0169 | Country | Pakistan | City | Karachi | Institution Type | University | Journal Type | Open Access | Manuscript Processing | Blind Peer Reviewed | Format | Paper Link | https://jict.ilmauniversity.edu.pk/journal/jict/18.1/7.pdf | Page |