Making Smart Contract Classification Easier and More Effective

Abstract

Nowadays, as an emerging and powerful technology, smart contract supports decentralized trusted computing and is widely used in various business fields such as finance, exchange and game. Although smart contracts provide a convenient way to manage funds in the business, they also bring economic loss to users. Particularly, there are a huge number of malicious smart contracts running on blockchain, which defraud users of money. Therefore, identifying the intention of smart contracts is of great significance, which will help users have a better understanding of smart contracts and reduce their financial losses. In this paper, we purpose a supervised classification method to detect the intention of smart contracts. Specifically, we extract the features of smart contracts from codes and transactions, and then use XGBoost to classify smart contracts. We conduct extensive experiments on on-chain smart contracts. The experimental results demonstrate that our features and model are capable of effectively identifying the types of smart contracts.