mwritescode/slither-audited-smart-contracts

Dataset Overview

Dataset Name

• Name: Slither Audited Smart Contracts
• Alias: Slither Audited Smart Contracts Dataset

Basic Information

• Language: English
• License: MIT
• Multilinguality: Monolingual
• Size: 100K < n < 1M
• Source: Raw data
• Task Types: Text classification, Text generation
• Task IDs: Multi‑label classification, Multi‑input text classification, Language modeling

Description

• Overview: The dataset contains source code and deployed bytecode of Solidity smart contracts verified by the Slither static analysis framework, along with vulnerabilities classified by Slither.
• Supported Tasks:
  • Text classification: Train models for binary and multi‑label classification of contract bytecode and source code.
  • Text generation: Train language models for the Solidity programming language.
  • Image classification: Convert bytecode to RGB images for CNN‑based vulnerability detection.
• Language: Annotations are in English; all source code is Solidity.

Structure

• Instances: Each instance includes address, source code, and bytecode. Labels are provided in two configurations: a cleaned textual version of Slither output, and a multi‑label version consisting of integer lists representing specific vulnerability classes. Label 4 indicates contract safety.
• Fields:
  • address: string, Ethereum main‑net contract address.
  • source_code: flattened Solidity code.
  • bytecode: string, contract bytecode obtained via web3.eth.getCode().
  • slither: cleaned JSON output from Slither or list of class labels.
• Splits: Six configurations exist; those without the all- prefix provide train, test, and validation splits, with test and validation each comprising ~15 % of the total data.

Creation

• Rationale: Provide a large, free dataset for detecting and classifying verified Solidity contract vulnerabilities.
• Source: Built from verified contract lists provided by Smart Contract Sanctuary; source code and bytecode were downloaded from Etherscan using Web3.py and analyzed with Slither.

Additional Information

• Creator: Martina Rossini
• License: Except for the Solidity source code, all files are under the MIT license.
• Citation: If used in research, cite as follows:

@misc{rossini2022slitherauditedcontracts,
    title = {Slither Audited Smart Contracts Dataset},
    author={Martina Rossini},
    year={2022}
}