Smart contract vulnerability detection

We propose a deep learning-based dataset for smart contract vulnerability detection, combining three public datasets to support and facilitate blockchain security research. This comprehensive dataset includes a variety of common types of smart contract vulnerabilities, such as re-entrancy attacks, integer overflows, and improper access controls.

By consolidating and uniformly annotating the data, we provide detailed vulnerability information and classification tags for each smart contract. The main features and contributions of the dataset are as follows:

The dataset includes smart contracts from three public datasets, ensuring diversity and representativeness of data sources.

Each smart contract is carefully annotated and details the vulnerability type, location, and associated code snippets for in-depth analysis by researchers.

The dataset contains not only the source code of the smart contract, but also bytecode and other related metadata, which can be analyzed and detected using a variety of techniques.

It is designed for training and evaluating deep learning-based vulnerability detection models. We provide standardized data formats to facilitate data preprocessing and model training by researchers.

It covers a wide range of common vulnerability types, ensuring that the dataset is comprehensive and useful.

We believe that this dataset will be an important resource for smart contract security research, especially deep learning-based vulnerability detection research. We hope that our work will advance the field and improve the security of smart contracts.

The dataset is linked as follows:

https://huggingface.co/datasets/mwritescode/slither-audited-smart-contracts

https://github.com/LaoSuanNaii/DL4SC

https://github.com/ZZXLX/contract_hex/tree/master