A Machine Learning-Based Dynamic Method for Detecting Vulnerabilities in Smart Contracts

Real-world application development through Smart Contracts on the Ethereum Blockchain platform is one of the emerging technologies. It also has much vulnerability, and reentrancy is among the most popular ones. In our work, we have reviewed the tools based on ML for vulnerability detection in Ethereum smart contracts. Based on that, we proposed a framework that can dynamically monitor threats based on the blockchain platform's transaction meta-data and balance data. It does not require any changes or updates to the existing system and does not require expertise to implement. This framework will extract features for machine learning classifier models from the transaction data and identify the transaction as agreeable or unfavorable. It will help to identify the reentrancy threat as well as the cause of it and help the developer to trace it from where the attack is generated. In the ML classifier for the framework, random forest and decision tree are used. The cumulative performance of both is 98 percent on 540 transactions.