A Deep Learning Model for Threat Hunting in Ethereum Blockchain

Abstract

Blockchain technology has found extensive applications in recent years, especially in financial and currency exchange applications, due to improved trustworthiness and security. Although blockchain technology improves security by design, it is not immune to security threats and vulnerabilities. Ethereum, as a decentralized, open-source blockchain, has shown high growth and widespread adoption in recent years, however, there is a wide range of vulnerability, security risks, and also attacks around it. To tackle such issues, machine learning could be a viable solution for threat hunting in the Ethereum blockchain. Machine learning algorithms, by analyzing the behav-ioral patterns, can achieve an insight for threat hunting. In this paper, we proposed a deep learning-based model for Ethereum threat hunting. The model applies a deep neural network for attack detection and uses a combination of machine learning algorithms (unsupervised with supervised algorithms) for attack classification. The performance evolution of the proposed model in terms of accuracy presents 97.72 % in Ethereum attack detection and 99.4% in attack classification.