The Onchainscan project developed a machine learning-based approach for detecting vulnerabilities in smart contracts, leveraging features from Opcodes and Control Flow Graphs (CFGs).
A standardised dataset was created across Solidity versions using AST-based bug injection and SWC classification, and it was published on the IEEE DataPort. The proposed MLP model achieved high accuracy while maintaining the lowest average False Positive Rate (FPR) among all evaluated tools and models, measured at 0.0125.
Project Deliverables/Outcomes/Impact
- Onchainscan prototype, 1 publication in IEEE Access
- Dataset published to IEEE data port
- Project showcased in HTX Tech Xplore exhibit
![[FA] SIT One SITizen Alumni Initiative_Web banner_1244px x 688px.jpg](/sites/default/files/2024-12/%5BFA%5D%20%20SIT%20One%20SITizen%20Alumni%20Initiative_Web%20banner_1244px%20x%20688px.jpg)
