ContractWard: Automated Vulnerability Detection Models for Ethereum Smart Contracts

Smart contracts are decentralized applications running on Blockchain. A very large number of smart has been deployed Ethereum. Meanwhile, security flaws have led to huge pecuniary losses and destroyed the ecological stability contract layer It is thus an emerging yet crucial issue effectively efficiently detect vulnerabilities in contracts. Existing detection methods like Oyente Securify mainly based symbolic execution or analysis. These time-consuming, as requires exploration all executable paths analysis dependency graphs a contract. In this work, we propose ContractWard with machine learning techniques. First, extract bigram features from simplified operation codes Second, employ five algorithms two sampling build models. evaluated 49502 real-world The experimental results demonstrate effectiveness efficiency ContractWard. predictive Micro-F1 Macro-F1 over 96% average time 4 seconds each when use XGBoost for training models SMOTETomek balancing sets.