SCsVulSegLyzer: Detecting and Extracting Vulnerable Segments from Smart Contracts Using Weakly-Supervised Learning
Smart contracts (SCs) are widely used in finance but remain attractive targets for hackers due to vulnerabilities, the immaturity of Solidity, and blockchain’s immutability. We present SCsVulSegLytix, a Transformer-based model that detects and extracts vulnerable segments directly from Solidity code using only contract-level labels, avoiding the need for costly graph-based models or line-level annotations. Supporting multiple vulnerability classes, it achieves higher efficiency and outperforms existing methods in both contract- and line-level vulnerability detection.