DIVE: A Multi-Label Smart Contract Vulnerability Dataset

📄 Dataset Description

The initial raw data was collected using the Etherscan API, leveraging three endpoints to gather:

🧩 Contract-based metadata

🧾 Account-based information

⚙️ Opcodes

The original dataset contained 269 attributes across thousands of Ethereum smart contracts. After applying multiple preprocessing steps, the final dataset includes:

238 features

22,330 samples

🏷️ Multi-Label Vulnerability Annotation

Each contract may exhibit multiple vulnerabilities, making this a multi-label dataset. Labels were generated using the MultiTagging framework, which analyzes each smart contract's source code using a suite of six tools: MAIAN, Mythril, Semgrep, Slither, Solhint, and VeriSmart. All tools were used with consistent versions and configurations, as specified in the MultiTagging project.

🛡️ Vulnerability Labels (DASP Top 10 Categories)

The dataset includes labels mapped to the first 8 categories of the DASP Top 10 vulnerability taxonomy:

Reentrancy

Access Control

Arithmetic Issues (Integer Overflow/Underflow)

Unchecked Call Return Values

Denial of Service (DoS)

Bad Randomness

Front Running

Time Manipulation (Timestamp Dependence)

🎯 Applications

This dataset is primarily designed for security analysis and machine learning tasks related to smart contract vulnerability detection. However, due to its rich feature set and multi-label structure, it may also be valuable for other research areas such as contract behavior modeling, feature interpretability, or anomaly detection.

📦 Package Contents

DIVE_*.csv: Final preprocessed dataset containing 240 features and 22,300 labeled smart contract samples.

DIVE_*_CategoricalColsMappings_*.json: Mappings for categorical feature values used during preprocessing.

DIVE_Labels_Vote_Result.csv: Multi-label vulnerability annotations generated using the MultiTagging framework.

Feature list.xlsx: Structured list of all features with names, types, descriptions, and category classifications.

EDA_and_Profiling_Reports.zip: Exploratory data analysis outputs with profiling reports and feature distribution plots.