Securing Smart Contracts in Fog Computing: Machine Learning-Based Attack Detection for Registration and Resource Access Granting

Abstract - Attack resistance in smart contracts has become one of the paramount concerns in this fog computing evolving landscape. Based on this problem, this study utilizes machine learning for the detection of attacks from the analysis of Ethereum transaction data and smart contract interactions. There are different methods for feature extraction: Term Frequency-Inverse Document Frequency (TF-IDF), Bag of Words (BoW), and N-gram methods for converting raw data into readable formats by machine. Several machine learning classifiers, including XGBoost, Random Forest, Light Gradient Boosting, and Extra Trees, are applied to detect the attacks. These models should offer high detection accuracy with low computational complexity, enabling scalability for realtime fog computing applications. Experimental results show that Extra Trees, with N-gram features, outperformed other models by achieving an accuracy of 83%. The proposed system is promising toward the efficient detection and classification of attacks in a dynamic fog environment.