Scholarlink

<p>This research investigates&#xa0;how deep learning can be used to detect fraud in Ethereum transactions on the blockchain. The Kaggle website (2022) is used to detect fraudulent activities. The data is quantitatively analyzed using the Chi-square statistical technique to identify the most relevant features for the model’s input. A graphical neural network is then developed, and performance evaluation metrics are employed to evaluate the effectiveness of the proposed detection model. Moreover, the study compares the performance of the newly proposed model with previous models in fraud detection. The findings indicate that among all classifiers, XGBoost achieves the highest accuracy levels, with a percentage of 98.34% when applied to the provided dataset. It is followed by RF with an accuracy of 97.88%, and ANN with an accuracy of 96.31%. On the other hand, LR demonstrates the lowest accuracy levels at 63.82%. The study concludes that the newly proposed model, based on deep learning techniques, is effective in detecting fraudulent activities and surpasses the performance of previous models in fraud detection. It also suggests further research to explore the application of other deep learning models in predicting fraudulent activities. Moreover, this study holds implications for researchers and decision-makers, emphasizing the potential of applying Ethereum cryptocurrency to deter criminals from engaging in fraudulent activities.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep learning techniques used for the detection of fraudulent activities within the ethereum network

  • 期刊论文

  • 发表日期：

    2025年10月27日

查看全文
  • Khalid Alkhatib,
  • Fares Al Ayoub,
  • Qais Marji

摘要

This research investigates how deep learning can be used to detect fraud in Ethereum transactions on the blockchain. The Kaggle website (2022) is used to detect fraudulent activities. The data is quantitatively analyzed using the Chi-square statistical technique to identify the most relevant features for the model’s input. A graphical neural network is then developed, and performance evaluation metrics are employed to evaluate the effectiveness of the proposed detection model. Moreover, the study compares the performance of the newly proposed model with previous models in fraud detection. The findings indicate that among all classifiers, XGBoost achieves the highest accuracy levels, with a percentage of 98.34% when applied to the provided dataset. It is followed by RF with an accuracy of 97.88%, and ANN with an accuracy of 96.31%. On the other hand, LR demonstrates the lowest accuracy levels at 63.82%. The study concludes that the newly proposed model, based on deep learning techniques, is effective in detecting fraudulent activities and surpasses the performance of previous models in fraud detection. It also suggests further research to explore the application of other deep learning models in predicting fraudulent activities. Moreover, this study holds implications for researchers and decision-makers, emphasizing the potential of applying Ethereum cryptocurrency to deter criminals from engaging in fraudulent activities.