Scholarlink

<p>Smart contracts are becoming more important to control operations/ transactions, particularly in fog computing environments. However, smart contracts can compromise security in several ways, especially the procedure that is used to give and register resource access. Since attackers might utilize sophisticated techniques to avoid detection, identifying these attacks can be a challenging task. This study provides a machine learning-based method to identify threats related to smart contracts in fog computing environments. Bidirectional encoder representations from transformers (BERT) and N-gram approaches are utilized for feature extraction and information gain with different threshold values of 0.01, 0.01, and 0.05 is employed for feature selection. In addition, data balancing is handled using the synthetic minority oversampling technique (SMOTE) and adaptive synthetic sampling (ADASYN). Alongside machine learning classifiers, deep learning models like convolutional neural networks (CNN) and graph neural networks (GNN) are also employed. According to experimental data, N-gram maintains its competitiveness with an accuracy score of 0.94, while BERT with ADASYN achieves the best accuracy score of 0.95. In terms of accuracy and F1 score, DistilBERT and MobileBERT outperform TinyBERT among deep learning models. These results highlight how hybrid machine learning approaches may be used to secure smart contracts in fog computing environments. Future research can explore real-time attack mitigation techniques, blockchain-based anomaly detection, and adversarial machine learning defenses.</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

Enhanced security of smart contracts in fog computing: hybrid classifiers and feature selection for improved attack detection for registration and resource access granting

  • 期刊论文

  • 发表日期：

    2025年9月11日

查看全文
  • Tahmina Ehsan,
  • Muhammad Usman Sana,
  • Alvena Ehsan,
  • Mustabeen Aziz,
  • Tahir Khurshaid,
  • Nagwan Abdel Samee,
  • Imran Ashraf

摘要

Smart contracts are becoming more important to control operations/ transactions, particularly in fog computing environments. However, smart contracts can compromise security in several ways, especially the procedure that is used to give and register resource access. Since attackers might utilize sophisticated techniques to avoid detection, identifying these attacks can be a challenging task. This study provides a machine learning-based method to identify threats related to smart contracts in fog computing environments. Bidirectional encoder representations from transformers (BERT) and N-gram approaches are utilized for feature extraction and information gain with different threshold values of 0.01, 0.01, and 0.05 is employed for feature selection. In addition, data balancing is handled using the synthetic minority oversampling technique (SMOTE) and adaptive synthetic sampling (ADASYN). Alongside machine learning classifiers, deep learning models like convolutional neural networks (CNN) and graph neural networks (GNN) are also employed. According to experimental data, N-gram maintains its competitiveness with an accuracy score of 0.94, while BERT with ADASYN achieves the best accuracy score of 0.95. In terms of accuracy and F1 score, DistilBERT and MobileBERT outperform TinyBERT among deep learning models. These results highlight how hybrid machine learning approaches may be used to secure smart contracts in fog computing environments. Future research can explore real-time attack mitigation techniques, blockchain-based anomaly detection, and adversarial machine learning defenses.