Deep learning-based methodology for vulnerability detection in smart contracts

Abstract

Smart contracts play an essential role in the handling and management of digital assets, where vulnerabilities can lead to severe security issues and financial losses. Current detection techniques are largely limited to identifying single vulnerabilities and lack comprehensive identification capabilities for multiple vulnerabilities that may coexist in smart contracts. To address this challenge, we propose a novel multi-label vulnerability detection model that integrates extractive summarization methods with deep learning, referred to as Ext-ttg. The model begins by preprocessing the data using an extractive summarization approach, followed by the deployment of a custom-built deep learning model to detect vulnerabilities in smart contracts. Experimental results demonstrate that our method achieves commendable performance across various metrics, establishing the effectiveness of the proposed approach in the multi-vulnerability detection tasks within smart contracts.

Keywords

• Smart contracts
• Vulnerability detection
• Multi-label classification
• Extractive sum-marization