Secure and Scalable Blockchain-Based Federated Learning for Cryptocurrency Fraud Detection: A Systematic Review

Abstract

Cryptocurrency fraud has become a significant threat in the digital financial ecosystem, with traditional detection systems often relying on centralized datasets and conventional machine learning techniques. These methods face challenges in terms of scalability, privacy, and the ability to effectively handle distributed and dynamic data sources. In this project, a secure and scalable approach is implemented for cryptocurrency fraud detection by leveraging a dataset stored on a blockchain. By utilizing the blockchain dataset, the system benefits from reliable and tamper-proof data while avoiding the need to implement blockchain infrastructure directly. Federated learning is employed to collaboratively train an XGBoost model across multiple data sources, enabling the detection of fraudulent transactions without centralizing sensitive information. This approach improves model accuracy, ensures data integrity, and provides a systematic framework for analyzing cryptocurrency transactions efficiently. The project demonstrates the potential of integrating advanced machine learning techniques with blockchainbased datasets for robust and privacy-preserving fraud detection.