Article: A hybrid deep learning-based framework for enhanced real-time fraud detection in Bitcoin transactions Journal: International Journal of Blockchains and Cryptocurrencies (IJBC) 2025 Vol.6 No.2 pp.89 - 112 Abstract: Fraud risks are on the rise with the increase in cryptocurrency transactions; the traditional detection methods become inadequate. This paper proposes a hybrid deep-learning framework for real-time fraud detection in Bitcoin transactions. Algorithms like Random Forest, support vector machine (SVM), Logistic Regression, and XGBoost are used to analyse transaction patterns and anomalies with a high level of accuracy. Different models have been tested in transaction data for Bitcoin, and the best-performing model was XGBoost with an accuracy of 96.94%. Advanced machine learning techniques enrich a system through secure data-driven insights and real-time anomaly detection, thus enhancing fraud risk detection. The obstacles faced are scalability, privacy issues, and inability to adjust models according to the evolving fraud technique. Future advancements might deal with federated learning, encryption methods, and cross-platform prevention to make the detection of fraud more secure. This paper indicates how well deep learning-based detection of fraud can scale and work efficiently to strengthen trust in digital financial systems. Inderscience Publishers - linking academia, business and industry through research

Title: A hybrid deep learning-based framework for enhanced real-time fraud detection in Bitcoin transactions

Authors: Sudip Diyasi; Ankita Ghosh; Dipankar Dey

Addresses: Department of Computer Application, Global Institute of Science and Technology, Haldia, Purba Medinipur-721657, West Bengal, India ' Department of Computer Application, OmDayal Group of Institutions, Uluberia, Howrah-711316, West Bengal, India ' Department of Computer Application, Global Institute of Science and Technology, Haldia, Purba Medinipur-721657, West Bengal, India

Abstract: Fraud risks are on the rise with the increase in cryptocurrency transactions; the traditional detection methods become inadequate. This paper proposes a hybrid deep-learning framework for real-time fraud detection in Bitcoin transactions. Algorithms like Random Forest, support vector machine (SVM), Logistic Regression, and XGBoost are used to analyse transaction patterns and anomalies with a high level of accuracy. Different models have been tested in transaction data for Bitcoin, and the best-performing model was XGBoost with an accuracy of 96.94%. Advanced machine learning techniques enrich a system through secure data-driven insights and real-time anomaly detection, thus enhancing fraud risk detection. The obstacles faced are scalability, privacy issues, and inability to adjust models according to the evolving fraud technique. Future advancements might deal with federated learning, encryption methods, and cross-platform prevention to make the detection of fraud more secure. This paper indicates how well deep learning-based detection of fraud can scale and work efficiently to strengthen trust in digital financial systems.

Keywords: fraud detection; Bitcoin transactions; hybrid framework; deep learning; cryptocurrency transactions; anomaly detection.

DOI: 10.1504/IJBC.2025.147778

International Journal of Blockchains and Cryptocurrencies, 2025 Vol.6 No.2, pp.89 - 112

Received: 09 Oct 2024
Accepted: 27 Feb 2025
Published online: 31 Jul 2025 *

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

Title: A hybrid deep learning-based framework for enhanced real-time fraud detection in Bitcoin transactions

Keep up-to-date