Phishing Detection on Ethereum Network menggunakan Metode Machine Learning

Abstract

This study discusses phishing detection on the Ethereum network using machine learning methods, specifically Graph Convolutional Networks (GCNs) and Enhanced Graph Attention Networks (EGAT). The background of this research is based on the increasing number of phishing attacks in the blockchain ecosystem that can threaten the financial security of users. The research aims to analyze the incidence rate of phishing attacks and develop effective and efficient detection methods. The methodology includes data collection from Ethereum transactions and phishing activities, followed by feature extraction, machine learning model training, and evaluation using metrics such as accuracy, precision, recall, and F-score. The identified research gap is the lack of focus on early-stage phishing detection in the Ethereum network and the suboptimal performance of existing methods in recognizing complex transaction patterns. The results indicate that EGAT achieves an accuracy of 93.6%, outperforming GCNs, which reach 91.2%. The conclusion of this research is that the EGAT method is superior in detecting phishing activities, providing significant contributions to security in the Ethereum network.