Background: In this paper, we address the paramount issue of detecting fraudulent transactions within the Ethereum blockchain by applying machine learning algorithms. The significance of fraud detection lies in its contribution to preserving the security and integrity of the blockchain, which is imperative for preventing monetary losses attributable to fraudulent activities. The dataset employed in this study is a publicly accessible compendium of Ethereum transactions, encompassing 9,841 transactions. Each transaction in the dataset is characterized by attributes, including gas price, transaction fee, and timestamp. Methods: The research paper is methodically divided into two primary sections: data preprocessing and predictive modeling. The dataset is accurately processed in the data preprocessing phase, and salient features are extracted from the transactions. This step is crucial in refining the data for optimal performance in the subsequent phase. Findings: In the predictive modeling phase, machine learning algorithms are deployed to classify transactions as fraudulent or legitimate. The algorithms evaluated include decision trees, logistic regression, gradient boosting, XGBoost, and a hybrid model integrating random forests with deep neural networks (DNN). Novelty: The findings suggest that the proposed model achieves a precision rate of 97.16%, indicating a significant enhancement in detecting fraudulent Ethereum transactions compared to the current leading techniques. This research contributes to the ongoing efforts to enhance the security and reliability of blockchain transactions through advanced analytical techniques..

Machine Learning algorithms, Blockchain, Deep learning, Fraud Detection, Ethereum