A comprehensive review of significant learning for anomalous transaction detection using a machine learning method in a decentralized blockchain network

Blockchain is a distributed ledger technology (DLT) that enables decentralized applications (DApps) such as Hyperledger, Bitcoin, Ethereum, decentralized finance (DeFi), and non-fungible token (NFT) to operate on it. Bitcoin is a popular application that leverages blockchain technology (BT) to provide secure transactions in a decentralized environment. Among the main reasons for BT being utilized is to eliminate the dependence of the process on third parties such as lawyers, insurance firms, and banks to execute approvals, verifications, and signatures. Due to the immutability and transparency of this technology, it has been deployed in several industries, including agri-food, supply chain, automotive, pharma, healthcare, insurance, land registration, and higher education, to increase verification, security, and traceability. Modern research indicates that blockchain distributed ledger may still be susceptible to various privacy, security, and dependability challenges, despite their impressive characteristics. To resolve these challenges, it is essential to notice abnormal behavior in a timely manner. Consequently, machine learning (ML) approaches with anomaly detection play a comprehensive role in preventing fraud. This paper explores the technological integration of anomaly detection models into BT and compares supervised and unsupervised ML techniques for detecting rogue and legitimate transactions. The studies for the review come from three well-known databases: Web of Science (WoS), Google Scholar and Scopus. Using sophisticated search tactics, specialized keywords such as Bitcoin, Ethereum, blockchain, fraud detection, anomaly detection, data mining, and ML are employed. These findings are based on three main points (1) the background of BT (2) Blockchain integration with ML, and (3) the ML approach for supervised learning and unsupervised learning. According to the findings of this study, supervised learning is the most prevalent method for studying anomalous detection in blockchain networks. This study also equips researchers with the knowledge necessary to perform research on the detection of blockchain network anomalies using ML techniques.