Phishing Attack Detection in Ethereum Transactions with PCA-Enhanced Machine Learning

Abstract

Phishing attacks in Ethereum transactions pose a significant threat to the security and integrity of blockchain-based systems, as these scams exploit user vulnerabilities to extract sensitive information or cryptocurrency assets. In contrast to the approaches proposed by the various research works to tackle phishing detection, many struggle with high-dimensional datasets, leading to computational inefficiencies and overfitting. To address these gaps, this study applies principal component analysis (PCA) for dimensionality reduction, helping in the development of more efficient and robust machine learning models by reducing data complexity and enhancing model generalization. A comparative analysis is conducted using multiple algorithms, including support vector machines (SVMs), decision trees (DT), XGBoost, and multi-layer perceptron (MLP). By evaluating their performance using standard metrics such as accuracy, F1 score, precision, recall, and ROC-AUC, the MLP model demonstrates superior accuracy and generalization, establishing its efficacy for phishing detection in Ethereum transactions. This work highlights the importance of feature reduction techniques and neural network models in enhancing the accuracy and efficiency of phishing detection systems, paving the way for future advancements in blockchain security.