Performance Evaluation of Volunteer Computing-based Neural Network on Ethereum Platform

In recent years, the application of distributed computing has become more and more widespread. Volunteer computing has gained people's favor in terms of only requiring low cost to exchange for a large amount of computing power. Volunteer computing is defined as a kind of distributed computing infrastructure with a large number of computers voluntarily provided by the public. The integrated computing and storage resources are collected and processed by a central server. However, for volunteer computing, security issues are of paramount importance due to the computing source provided by the volunteers. The attack on any computer in the volunteer network may have an impact on the whole network. On the other hand, blockchain is a chain storage structure that has been widely used in recent years, and non-tamperability is one of its most important characteristics. Therefore, this research aims to improve the security and credibility of volunteer computing by using blockchain technology. Accordingly, a novel volunteer network is established on the Ethereum test network named Ropsten. Its performance and consumption are evaluated with a neural network used for computing. The results demonstrate that proper security properties of volunteer computing can be guaranteed by the proposed blockchain-based approach, although a comparatively low time cost and commission consumption is required.