IDENTIFICATION OF FRAUDULENT TRANSACTIONS ON ETHEREUM BLOCKCHAIN

Decentralization, transparency, and immutability are the three pillars on which the blockchain’s digital system is built. Two well-known Blockchain platforms, Bitcoin and Ethereum, allow users to deal anonymously online while being globally connected peer-to-peer. The enormous volume of decentralized transactions and the participants’ apparent anonymity make it possible for scams, cyber frauds, hacks, money laundering, and fraudulent transactions. Since standard auditing approaches require more processing power, time, and memory for complicated queries to join combinations of data, it is difficult to identify such fraudulent acts using them. Utilizing machine learning algorithms is one strategy to get rid of fraud. Machine learning can be either supervised or unsupervised in nature. Several machine learning strategies have been presented in earlier works to address this issue, and while some of the results seem extremely promising, there is no one approach that stands out as being clearly superior. In those studies, the effectiveness of several supervised machine learning models, including SVM, Decision Tree, Naive Bayes, Logistic Regression, and a few deep learning models, was compared in order to identify fraudulent transactions in a blockchain network.We notice that these models have limitations such as scalability and accuracy. We propose a new Machine Learning method i.e, XGB classifier to handle the above problem in addition to existing methods. Where we discuss the limitations like scalability, accuracy and handling of missing values. We conclude by highlighting the advantages of our proposed system in comparison to that of previous existing methods.