Abstract

An overall strategy is suggested on how blockchain security can be improved through advancement of machine-learning based fraud detection targeted at Ethereum transactions. The study utilizes different six machine learning algorithms such as Ada Boost Classifier, LGBM, Extra Trees Classifier, XGBoost, Random Forest Classifier and Gradient Boosting Classifier that analyze its accuracy in detecting fraudulent acts within ethereum networkUsing experiments and comparisons, it reveals that LGBM algorithm gives the maximum fraud detection performance in Ethereum transactions. Systematic use of each algorithm incorporating hyperparameter tuning for better models. Precision, recall, and F1-score were selected as the metrics for evaluating each model in order to provide a holistic analysis.The results stress the role of machine learning in strengthening transaction security on Ethereum, and the highlighted LGBM algorithm is very efficient. Hence, this study provides useful information for future development of enhanced security mechanisms for blockchain-based ethereum transactions.