Dataset and Reproducibility for ReSect: A Tool for Automated Analysis of Reentrancy Transactions on Blockchains

This dataset accompanies ReSect, a tool for automated post-mortem analysis of reentrancy transactions on EVM-compatible blockchains. It contains two PostgreSQL dumps used in the DSN 2026 evaluation: A smaller attack-only dataset for false-negative evaluation, containing real-world reentrancy exploit transactions from 73 attacks (1,122 exploit transactions in total). A full evaluation dataset for false-positive evaluation, containing 1,077,298 non-trivial randomly sampled transactions. The first dataset is a subset of the second and is provided separately to support validators with limited time, storage, or hardware; validators with sufficient resources may download and restore only the full dataset and run both false-negative and false-positive evaluations from it. The data are preprocessed and cached to support reproducible experiments without external API access, enabling validators and researchers to reproduce ReSect’s accuracy and large-scale false-positive evaluation. For detailed instructions, please see README.md.本数据集配套ReSect工具使用，该工具用于对兼容EVM（Ethereum Virtual Machine，以太坊虚拟机）的区块链上的重入交易开展自动化事后分析。本数据集包含用于DSN 2026评估的两份PostgreSQL数据库备份文件： 其一为仅针对攻击场景的小型数据集，用于假阴性评估，收录了来自73起真实重入漏洞攻击的共计1122笔攻击交易。 其二为用于假阳性评估的全量评估数据集，收录了1077298笔经随机采样的非平凡交易。 首个数据集为第二个数据集的子集，单独发布以帮助受限于时间、存储或硬件资源的验证人员；资源充足的验证人员可仅下载并恢复全量数据集，即可通过该数据集完成假阴性与假阳性两类评估。 本数据集已完成预处理与缓存，支持无需依赖外部API的可复现实验，使验证人员与研究人员能够复现ReSect的准确率测试与大规模假阳性评估流程。详细操作指南请参见README.md文件。