Fraud Transactions of Ethereum Smart Contract Using Modified Whale Optimization Algorithm with Deep learning

Anomaly detection in smart contracts is an effective means of preventing concealed security risks like illicit financing financial fraud, and money laundering. Currently, Ethereum stands as the leading platform for smart contracts, making anomaly detection a pressing concern. Nonetheless, the voluminous and intricate nature of smart contract data renders the extraction of highlevel attributes difficult and traditional methods inefficient. Thus, this study proposed a deep learning model for detecting fraudulent transactions on the Ethereum blockchain. The model utilizes a Deep Neural Network (DNN) with 45 concealed layers and is optimized using the modified Whale Optimization Algorithm (WOA) with a Gaussian chaotic map. The model's performance is evaluated using various performance metrics, and the results are presented in a confusion matrix and ROC curve. With an accuracy rate of 99.78%, the proposed approach surpasses conventional machine learning approaches such as Support Vector Classification (SVC), k-Nearest Neighbors (KNN), Logistic Regression (LR) and Random Forest (RF), and along with other well-known deep learning models. Experimental results demonstrates the potential of proposed model in detecting fraudulent activities and securing decentralized applications.