Abstract The security of smart contracts is vital to the stability of the blockchain ecosystem. These contracts are self-executing programs deployed on the blockchain. Traditional fuzz testing often suffers from limited coverage and low efficiency. It struggles to trigger boundary conditions and explore complex contract states. To address these challenges, this paper proposes AMFUZZ, a boundary-aware and multi-strategy adaptive mutation-based fuzz testing framework. AMFUZZ enhances test case generation and state space exploration. It adjusts mutation amplitude dynamically and applies exponential growth strategies. These techniques modify transaction values based on input scale and contextual information. The goal is to improve the chances of reaching boundary conditions. AMFUZZ also introduces multiple mutation modes. It combines large and small mutations flexibly during testing. Furthermore, it uses parameter combination analysis to improve execution efficiency. Experimental results show that AMFUZZ improves code coverage and transaction throughput, while reducing memory usage and total execution time in multiple benchmark contract tests.
Zhang et al. (Mon,) studied this question.