Research On Smart Contract Vulnerability Detection Technology Based On Deep Learning

In recent years,blockchain technology has attracted more and more attention in finance,medical and other industries.Blockchain and cryptocurrency are gaining unprecedented popularity and welcome.Smart contract is an important part of blockchain technology.With the development of blockchain,the security of smart contract has become a hot research field.At the same time,as an open-source public blockchain platform with smart contract function,Ethereum has also developed rapidly.The design method of Ethereum enables developers to write smart contracts and distributed applications more efficiently.Ethereum platform provides the possibility and opportunity for the rapid development of blockchain technology in all walks of life.However,the security of Ethereum smart contract has not received much attention.From2011 to now,the economic losses caused by smart contract loopholes have reached tens of billions of dollars.There are many kinds of smart contract vulnerabilities.The common ones are transaction order dependency,time stamp dependency,re-entry vulnerability and improper exception handling.This paper analyzes the causes and effects of each vulnerability through modeling,and introduces deep learning technology to detect smart contract vulnerabilities.In the face of various types of loopholes in the smart contract,the traditional vulnerability detection method uses static and dynamic analysis to find the loopholes in the smart contract.Its audit process is low in automation and high in false alarm rate.Sometimes it needs manual secondary audit.In this paper,we focus on reducing the labor cost of technicians and improving the efficiency of smart contract vulnerability detection,and propose a framework of smart contract vulnerability detection based on deep learning.The framework is divided into preprocessing module,model training module and vulnerability detection module.Firstly,the smart contract is input into the data preprocessing module and transformed into a vector suitable for neural network input.Then,the vector is input into the model training module to train the neural network model with the optimal parameter combination.Finally,the trained model is put into the vulnerability detection module todetect the vulnerability of the smart contract and output the vulnerability category.Through the use of deep learning technology,this paper improves the accuracy and time efficiency of vulnerability detection of smart contracts,greatly reducing the burden of blockchain technicians.