Ethereum, a well-known blockchain's most famous implementation builds and deploys a decentralized application where users can transact with cryptocurrency using smart contracts

Abstract

Ethereum, a well-known blockchain’s most famous implementation builds and deploys a decentralized application where users can transact with cryptocurrency using smart contracts. The latest technological developments in cryptocurrencies and the benefits associated with them have been hidden by a number of illegal activities on the network. Like bribery, phishing scam, money laundering and fraud etc. The ‘Pseudoanonymous’ nature of the participants in Ethereum blockchain network leads to cause difficulty in detecting illicit behavior of the users. Anomalies must be detected and resolved quickly to ensure participant trust in the largest blockchain platforms. There is a lot of work on detailed analysis of Ethereum transactions in terms of how well they work, but this research is the first to the best of my knowledge to detect anomalies in Ethereum transaction records. To achieve our goal, we have extracted more comprehensive feature of Ethereum transaction data to get rid of the shortcomings of existing work. We have used SMOTE technique to deal with highly imbalanced dataset and implemented five supervised-machine-learning model, Logistic-Regression, KNN, Decision-tree, Random-Forest and SVC classifier to access and compare the best performer among them. Random Forest Outperforms with accuracy of 98%. We evaluated the accuracy, precision, and F-value of each method and backed them up with experimental results.