A Feature-Based Robust Method for Abnormal Contracts Detection in Ethereum Blockchain

Blockchain technology has allowed many abnormal schemes to hide behind smart contracts. This causes serious financial losses, which adversely affects the blockchain. Machine learning mainly been utilized enable automatic detection of contract accounts in recent years. In spite this, previous machine methods have suffered from a number disadvantages: first, it is extremely difficult identify features that accurate contracts, and based on these features, statistical analysis also ineffective. Second, they ignore imbalances repeatability accounts, often results overfitting model. this paper, we propose data-driven robust method for detecting over Ethereum Blockchain. comprises hybrid set by integrating opcode n-grams, transaction term frequency-inverse document frequency source code train an ensemble classifier. The extra-trees gradient boosting algorithms weighted soft voting are used create classifier balances weaknesses individual classifiers given dataset. normal data collected analyzing open etherscan.io, problem imbalanced dataset solved performing adaptive synthetic sampling. empirical demonstrate proposed feature sets useful accounts. Meanwhile, combining all enhances contracts with significant accuracy. experimental comparative show can distinguish security blockchain satisfactory performance metrics.