Smart Contract Fuzzing Based on Reinforcement Learning
With the rapid development of blockchain technology,smart contracts are increasingly widely used in digital asset transactions and other fields.However,the problem of its security vulnerability is also becoming more and more prominent,posing a serious threat to the security of the blockchain system.In view of this,a smart contract fuzzing method based on reinforcement learning,RL_soFuzzer,is proposed to improve the security detection efficiency for smart contracts.Conventional fuzzing methods require a large number of transaction sequences to trigger potential vulnerabilities,resulting in state space explosion and poor code coverage.RL_soFuzzer uses reinforcement learning technology to schedule memory snapshots,and combines data stream correlation sorting for in-depth detection.At the same time,the function state variable sorting mechanism based on topological sorting is introduced to reduce the uncertainty in the testing process.Experimental results indicate that the RL_soFuzzer is significantly better than conventional methods in terms of code coverage and vulnerability detection ability,and shows excellent performance in practical applications.
NETL
NSTL
万方数据
