Sol2Vy : Leveraging Solidity-Trained Models for Vyper Smart Contract Analysis

Minghao Hu, Qiang Zeng, Lannan Luo

Keywords: smart contract, vulnerability detection, intermediate representation

TL;DR: We use SlithIR as a bridge to transfer knowledge from Solidity to Vyper for vulnerability detection

Abstract: Smart contracts have transformed decentralized finance, but flaws in their logic still create major security threats and financial losses. Most existing vulnerability detection techniques focus on well-supported languages like Solidity, while low-resource counterparts such as Vyper remain largely underexplored due to scarce analysis tools and limited labeled datasets. To address this gap, we introduce Sol2Vy, a novel framework that enables vulnerability detection in Vyper smart contracts using models trained solely on Solidity. The key enabler is SlithIR, which we leverage as an intermediate representation to achieve effective cross-language knowledge transfer. Our framework follows a principled three-stage design that integrates unsupervised knowledge learning, supervised vulnerability detection model training on Solidity, and cross-language testing on Vyper. This approach eliminates the need for extensive labeled Vyper datasets typically required to build an accurate vulnerability detection model. We implement and evaluate Sol2Vy on three critical vulnerability types: reentrancy, weak randomness, and unchecked transfer. Experimental results show that Sol2Vy achieves strong detection performance on Vyper contracts despite being trained exclusively on Solidity, significantly outperforming existing tools.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Submission Number: 3471