Smart Contract Datasets of sGuard+

The repository consists of three datasets: Unlabled Real-world Smart Contract Dataset (URD), Labled Real-world Smart Contract Dataset (LRD), Publicly Available Smart Contract Vulnerability Dataset (PVD), which are used to train machine learning models of sGuard+ and evaluate the effectiveness and efficiency of sGuard+. Labels in these datasets involve the following five vulnerabilities:

• SWC-101: Integer Overflow and Underflow Vulnerability (IOU)
• SWC-104: Unchecked Call Return Value Vulnerability (UCR)
• SWC-106: Unprotected SELFDESTRUCT Instruction Vulnerability (USI)
• SWC-107: Reentrancy Vulnerability (REN)
• SWC-115: Authorization through Tx-origin Vulnerability (TXO)

Unlabled Real-world Smart Contract Dataset (URD)

https://github.com/renardbebe/Smart-Contract-Benchmark-Suites/tree/master/dataset/UR

Labled Real-world Smart Contract Dataset (LRD)

We label URD by manually confirming the detection results of Slither, Securify (and v2.0) and Mythril.

Publicly Available Smart Contract Vulnerability Dataset (PVD)

We collect the vulnerable smart contracts from SWC Registry, CVE, and SmartBugs Curated as a ground truth of the five vulnerabilities mentioned above.

SCRepair Dataset (SCRD)

https://github.com/xiaoly8/SCRepair