ANOMALY DETECTION IN CRYPTOCURRENCY TRANSACTIONS USING MACHINE LEARNING

Abstract

Cryptocurrencies have enhanced financial transactions, but being decentralized, they pose numerous security threats to their users, warranting new anomaly detection systems for fraud prevention.The present research focuses on the machine learning (ML) techniques used in detecting suspicious activities in cryptocurrency networks, focusing on their contribution to AML and CFT compliance.The paper also compares supervised and unsupervised learning techniques and their merits and demerits.The supervised learning techniques, including Decision Trees, SVMs, and Neural Networks, are presented for their accuracy and flexibility, and, on the other hand, the unsupervised learning approaches, including Clustering, Isolation Forests, and Autoencoders are considered for their potential to discover new fraud patterns even if the training data is not labeled.An analysis of the use of explainability tools such as LIME and SHAP in artificial intelligence systems is also carried out to improve how users understand the results given to them by the AI models.These models have their real-life application illustrated by case studies, which prove helpful in identifying anomalies in Bitcoin and Ethereum transactions.New research directions suggest improvements in machine learning methods, the connection of the results with analysis tools based on blockchain, and cooperation with relevant authorities to improve the identification of threats and conformity with established guidelines.The potential of applying the idea of this work in traditional finance and cybersecurity is discussed, highlighting the possibility of applying ML in multiple fields to enhance security and compliance.The study then informs the significance of continued research and collaboration among disciplines to combat the emerging issues of financial fraud and cybercrimes related to cryptocurrencies.