SVM Machine Learning Model for Detection of Etherlock Vulnerability in Solidity Smart Contracts

Blockchain is a disruptive technology that changed how financial operations and ownership exist over the internet. With the implementation of smart contracts in blockchains, facilitation of transactions without intermediaries is possible and fully automated. Smart contracts allow relatively safe transactions and transfer of ownership, but as with any piece of software, it proved to have vulnerabilities and challenges. Unfortunately, considering that smart contracts hold and facilitate valuable asset and funds transfer, the damage caused by exploitation of these contracts is devastating with billions of dollars’ worth of cryptocurrency and other assets lost or stolen up until now. One attack on Ethereum smart contracts is the Etherlock attack, which exploits bad logic in the code to stall the transaction indefinitely, effectively sealing user funds. In this paper, we propose a high-accuracy Support Vector Machine model for detection of Etherlock vulnerability in Solidity smart contracts. The model is developed into a standalone tool and achieves 98.44% accuracy accompanied with 87% recall using a generated dataset of Smart Contracts from the Ethereum blockchain. Our promising results still show the necessity of more sophisticated accurate detection models for protection of crypto assets on the blockchain.