TOWARDS AUTOMATED SMART CONTRACTS VULNERABILITY DETECTION BASED ON DEEP LEARNING MODELS

Abstract

Smart contracts are at the core of today's strong development of blockchain technology. However, due to the immutability and publicity of smart contracts, potential vulnerabilities in them become extremely dangerous. The issue of finding vulnerabilities in smart contracts thus attracts a lot of attention and is a hot issue in the development of blockchain applications. This paper proposes a model for applying deep learning technology in detecting and classifying vulnerabilities in smart contracts. The proposed model is developed based on 1D CNN architecture. This model gives better performance than traditional 2D CNN models in the problem of detecting vulnerabilities in smart contracts. We collect and label smart contract vulnerabilities using the Slither engine and perform proposed model evaluation with several traditional CNN models. The results show that the proposed model has a higher performance, reaching an accuracy of 98.18%. The results show the applicability of deep learning technology in software vulnerability detection in general and smart contracts in particular. This is the basis for developing more secure smart contracts in practice.