Ethereum virtual machine(EVM) is the heart of the Ethereum infrastructure and functions as the runtime environment for smart contracts. Inconsistencies in EVMs may cause transaction sender’ financial losses. Furthermore, EVM’s vulnerabilities may lead to serious problems such as transaction failure or even client crashes. Thus, it is critical to ensure the reliability and consistency of mainstream EVM implementations. However, not much work has been done on validating EVMs. In this thesis, we design 5 mutators to continuously manipulate the control flow of seed bytecodes of smart contracts to generate semantically different mutants. These mutants execute on a reference EVM and use branch coverage of EVM as guidelines to selectively accept the generated mutants. This step steers...[ Read more ]

Ethereum virtual machine(EVM) is the heart of the Ethereum infrastructure and functions as the runtime environment for smart contracts. Inconsistencies in EVMs may cause transaction sender’ financial losses. Furthermore, EVM’s vulnerabilities may lead to serious problems such as transaction failure or even client crashes. Thus, it is critical to ensure the reliability and consistency of mainstream EVM implementations. However, not much work has been done on validating EVMs. In this thesis, we design 5 mutators to continuously manipulate the control flow of seed bytecodes of smart contracts to generate semantically different mutants. These mutants execute on a reference EVM and use branch coverage of EVM as guidelines to selectively accept the generated mutants. This step steers the mutation process toward diverse mutants. The generated mutants are then employed to differentially test EVMs.

We have evaluated this technique on mainstream EVM implementations, including go-ethereum, openethereum and aleth. Compared to existing work, the experimental results show that our approach can (1) generate more valid test cases in a short time period; (2) achieve higher code coverage in EVM and (3) discover some previously unknown issues among EVMs.