Gan-Enabled Code Embedding For Reentrant Vulnerabilities Detection

As one of the key components of blockchain, smart contract is playing a vital role in achieving auto-functions; however, reentrant attacks are threatening the implementation of smart contracts, which limits the adoption of blockchain systems in various scenarios. To address this issue, we propose a reentrant vulnerability detection model based on word embedding, similarity detection, and Generative Adversarial Networks (GAN). Additionally, we provide a new approach for dynamically preventing reentrant attacks. We also implement experiments to evaluate our model and results show our scheme achieves 92% detecting accuracy for reentrant attack detection.