Smart Contract Reentrancy Vulnerability Detection Method Based on Deep Learning Hybrid Model

Abstract

With the advancement of blockchain technology, smart contracts, as one of its core applications, have been widely utilized across various industry sectors. However, smart contracts face significant security challenges, with reentrancy vulnerability being a typical concern. In order to effectively detect reentrancy vulnerabilities in smart contracts, this paper proposes a parallel hybrid model based on deep learning. The proposed method initially employs Word2Vec model for word embedding, preprocessing, and feature vectorization of the smart contract code data. Subsequently, a combination of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) is employed for feature extraction and sequence modeling. Simultaneously, the method utilizes a parallel approach combining Recurrent Neural Network (RNN) and Gated Recurrent Unit (GRU) with CNN-LSTM to enhance the model's learning capabilities and efficiency in handling sequential data. Finally, a fully connected layer and a Softmax classifier are employed to classify the extracted features. Through a series of experiments and performance evaluations, the proposed method demonstrates significant improvements in key metrics such as precision, recall, and F1 score compared to traditional methods and single deep learning approaches, achieving 91.66%, 90.16%, and 90.90% respectively.